Paid media buys attention, organic content earns it, and 2026 changed the price of both

Meta delivered 19% more ad impressions in the first quarter of 2026 than in the same quarter of 2025, and charged 12% more for each one. Google Search and other advertising revenue grew 19% year over year to $60.4 billion in the same period, while Google Network revenue — the money Google makes placing ads on other people’s websites — fell 4% to $6.97 billion. Those three numbers, taken from the companies’ own quarterly filings, describe the situation more precisely than any agency deck. Bought attention is getting more expensive and more abundant at the same time. Earned attention is getting harder to route to a website you own.

Marketers who ask whether paid media or organic content is better are usually asking one of four different questions without separating them. They may be asking which channel produces cheaper incremental revenue this quarter. They may be asking which produces more durable revenue over three years. They may be asking where to put the next €10,000 given a fixed team. Or they may be asking which discipline their company should build internal capability around. Those four questions have four different answers, and the answers change depending on category, margin structure, sales cycle length, brand maturity, and the amount of capital available to wait.

The comparison is also unstable because the two things being compared are not the same kind of object. Paid media is a purchasing decision with a price, a delivery guarantee of sorts, and a measurable stop button. Organic content is a production and asset-building decision with a long, uncertain payback and no delivery guarantee at all. Comparing them directly is closer to comparing rent to a mortgage than comparing two rental properties. You can stop paying rent tomorrow and lose access immediately; you can stop paying the mortgage and still own equity, but you cannot get the equity quickly and you may have overpaid for the house.

What has genuinely changed by 2026 is not the logic of that comparison but the parameters inside it. Three shifts matter. First, the price of paid impressions has risen faster than most companies’ gross margins, because the number of advertisers bidding into automated auctions grew faster than the supply of high-intent attention. Second, the click-through economics of organic search weakened materially once generative summaries were inserted above the results — Pew Research Center found that users who saw an AI summary clicked a traditional result link in 8% of visits, against 15% when no summary appeared. Third, a new distribution surface appeared, in the form of assistants and answer engines, that behaves like organic search in that you cannot buy your way into it reliably, and behaves like paid media in that the platform intends to monetise it.

Anyone selling a clean answer is selling something. The defensible position is narrower and more useful: there is a correct ratio for a given business at a given moment, it is discoverable through experiment rather than argument, and it moves. This analysis is an attempt to specify what determines that ratio, what the 2026 evidence actually supports, where the evidence is thin, and how to run the decision inside a real company with a real budget.

The rest of this piece works through the mechanics of both systems, the measurement problems that make honest comparison difficult, what the current data supports and what it does not, sector-specific arithmetic, the regulatory and platform risks attached to each side, and a concrete operating procedure for setting and revising the split. Where a claim is interpretation rather than measurement, it is marked as such.

Definitions that actually matter inside a media plan

Loose vocabulary is responsible for a large share of bad media decisions. Before any comparison is possible, the categories need to be tight enough to be countable.

Paid media is any distribution where a platform or publisher delivers your message to an audience in exchange for money, priced per impression, per click, per view, per acquisition, or per unit of time. Search ads, social ads, retail media placements, connected TV, programmatic display, sponsored newsletter slots, and paid influencer contracts all sit here. The defining property is not that money changes hands — money always changes hands somewhere — but that distribution itself is the thing being purchased, and that distribution stops when payment stops.

Organic content is distribution earned through the value of the content itself, mediated by an algorithm or a human decision you do not control. Search results that rank on merit, social posts that reach a feed without promotion, YouTube videos surfaced by recommendation, newsletters opened because someone subscribed, citations inside an AI answer, mentions in a forum thread. The defining property is that no per-unit distribution fee is charged, though the content carries a real production cost and the distribution is conditional on someone else’s ranking logic.

Two categories sit awkwardly between them and cause most of the confusion. Owned media — your website, your email list, your app, your physical stores, your customer database — is not organic in the traffic-acquisition sense. It is the destination and the retention layer. Email in particular is often filed under organic when it behaves more like a distribution asset with near-zero marginal cost and high control. Earned media — press coverage, word of mouth, user-generated posts, review-site presence — is genuinely uncontrolled, and companies routinely take credit for it inside organic reporting when they had limited influence over it.

Two more terms need pinning down because the entire argument turns on them. Incrementality is the share of a measured outcome that would not have occurred without the activity. A conversion attributed to a branded search ad by an ad platform may have close to zero incrementality if the person would have clicked the organic result immediately below. Marginal cost per incremental acquisition is the number that should drive allocation: not average cost per acquisition across a channel, but the cost of the next customer from the next euro in that channel. Average CPA flatters mature paid campaigns and flatters organic almost always, because organic average CPA divides a fixed production cost by a growing denominator.

A further distinction matters in 2026: generative distribution. Content surfaced or cited inside an AI assistant answer is not classical organic search, because there is often no click, no session, and no analytics record. It resembles brand awareness delivered by a third party who does not tell you it happened. Some practitioners call the practice of pursuing it generative engine optimisation, or GEO; others fold it into answer engine optimisation. The vocabulary is unsettled, but the measurement problem it creates is concrete: a growing share of commercially useful exposure now happens in surfaces that produce no measurable visit.

Finally, treat CAC, blended CAC, and paid CAC as three separate numbers, because conflating them is how organic content gets credited with paid results and vice versa. Paid CAC divides paid spend by customers attributed to paid. Blended CAC divides total marketing and sales cost, including salaries and content production, by all new customers. Blended CAC is the honest number for board reporting. Paid CAC is the honest number for bid decisions. Neither answers the other’s question.

The budget war that produced the false choice

The paid-versus-organic framing is not an eternal feature of marketing. It emerged from a specific set of conditions between roughly 2009 and 2016, and understanding that origin explains why the framing survives past its usefulness.

Search engine optimisation and content marketing became a mass discipline during a window when Google’s index was comparatively easy to influence and paid search inventory was comparatively cheap. Companies could publish volume, rank, and receive free traffic that converted. The economics were startling: a blog post costing €300 could deliver traffic for years. Agencies built businesses on that arbitrage, and a generation of marketing leaders formed their instincts inside it. Content marketing’s reputation as a compounding asset was earned in an environment that no longer exists in the same form.

Paid social arrived with its own arbitrage. Facebook’s auction in the early 2010s was under-subscribed relative to the attention it had accumulated, and early advertisers bought reach at prices that later looked absurd. Direct-to-consumer brands were built almost entirely on that gap. When the gap closed — through advertiser density, then through Apple’s App Tracking Transparency changes in 2021, then through general auction maturity — many of those businesses discovered their unit economics had been a function of temporary cheap distribution rather than product advantage.

Both disciplines then hardened into professional identities, and that is the mechanism that keeps the false choice alive. An SEO lead and a paid media lead are not two analysts weighing the same evidence; they are two people whose budgets, headcount, and career progression depend on different answers. Agencies compound the problem, because most agencies are structurally better at one than the other and pitch accordingly. Vendor content follows the same incentive: measurement platforms that sell attribution favour arguments where attribution is decisive, and content platforms favour arguments where durable assets win.

Three further developments pushed the two disciplines into direct budget conflict rather than complementary roles. Organic reach on social platforms compressed as feeds shifted from chronological follower delivery to recommendation-driven ranking, which decoupled audience size from reach and made paid promotion the reliable route to distribution. Search results pages grew commercially denser, pushing organic listings further down. And marketing budgets stopped growing — Gartner’s 2025 CMO Spend Survey put marketing budgets at 7.7% of company revenue, described as flatlined, meaning almost every increase in one line item became a decrease somewhere else.

By 2026 a fourth pressure arrived: artificial intelligence started consuming budget as a line item in its own right. Gartner’s 2026 CMO Spend Survey, based on 401 CMOs and marketing leaders across North America, the UK and Europe surveyed between January and March 2026, found 15.3% of marketing budgets allocated to AI while only 30% of CMOs described their organisations as ready to scale AI capabilities. Labour rose to 24.5% of marketing budget in 2026 from 21.9% in 2025. The same research found awareness and conversion objectives together consuming 62.6% of total media spend, with loyalty and retention spending down 29% since 2024 and now under 15% of media.

That last figure deserves attention, because it describes a market steadily moving money toward the top and bottom of the funnel and away from the middle — and the middle is precisely where organic content historically did its most useful work.

Mechanics of paid media pricing

Understanding why paid media costs what it costs in 2026 requires understanding the auction, and most advertisers hold an outdated mental model of it.

Every major paid platform runs a real-time auction where the winning bid is not simply the highest monetary offer. Google’s ad rank multiplies bid by quality signals and expected impact from ad extensions and formats. Meta ranks on estimated action rate multiplied by bid, plus an ad quality term. The practical consequence is that the price you pay is set by other advertisers’ willingness to pay for the same person, adjusted by how well the platform expects your ad to perform. You are not buying inventory. You are buying priority in a contest for a specific human at a specific moment.

Three forces set the price level. The first is advertiser density: how many bidders want the same audience. The second is the platform’s ability to predict conversion, because better prediction lets the platform charge closer to the true value of the click. The third is inventory supply, which grows when platforms add surfaces — Reels, Stories, Shorts, Discover, search partner networks, retail media pages, connected TV pods.

The 2026 pattern in the disclosed data is that supply and price rose together, which is unusual and informative. Meta’s 19% impression growth alongside 12% price growth means the company found more places to put ads and still raised the clearing price. That combination happens when demand grows faster than supply, and when prediction quality improves enough that advertisers tolerate higher nominal prices because measured returns hold. Meta reported $56.31 billion in first-quarter 2026 revenue, up 33% year over year, with 3.56 billion daily active people across its family of apps in March 2026.

The prediction-quality point is the one most advertisers underweight. Automated campaign types — Performance Max, Advantage+, and their equivalents — moved control from the advertiser to the platform’s model. The trade is real: platforms with better signal find conversions the advertiser’s manual targeting would have missed, and in exchange the advertiser loses visibility into which audiences, placements and queries produced the result. The efficiency gain is genuine and the loss of diagnostic information is also genuine, and the second cost is invisible on a dashboard.

There is a structural asymmetry worth stating plainly. Platform automation is designed to find conversions at the advertiser’s stated target, not to find incremental conversions. If a model can hit a €30 cost-per-acquisition target by serving ads to people already intending to buy — retargeting site visitors, bidding on your own brand name, reaching existing customers — it will, because that path satisfies the objective at lowest cost. Automated bidding is, by construction, biased toward harvesting demand rather than creating it, and harvested demand is the demand most likely to have converted anyway. This is not a conspiracy; it is what happens when you optimise a proxy metric.

Pricing also varies enormously by intent depth and category. High-intent commercial queries in insurance, legal services, and B2B software routinely clear at costs per click in the tens of euros, because the value of one converted customer is in the thousands. Low-intent social impressions in consumer categories clear at a few euros per thousand. Neither number tells you anything on its own; the ratio between cost and customer lifetime value is the only figure that matters, and it is the figure most reporting hides.

Two further pricing dynamics changed the 2026 picture. Retail media expanded fast, adding inventory that sits closest to purchase and competes directly with lower-funnel search budgets. And the arrival of ads inside AI answer surfaces created a new, thin, expensive inventory class. Search Engine Land’s March 2026 analysis argued Google would monetise AI Mode lightly and gradually — even a small share of sessions carrying ads generates the data needed to build the format — while noting that eligibility flows from existing Performance Max, standard shopping, and keyword campaigns rather than a separate buying motion.

Inside the organic distribution machine

Organic distribution is often described as if it were one system. It is at least five, with different logics, different failure modes, and different 2026 trajectories.

Classical search ranking remains the most documented. A crawler discovers pages, an index stores them, and a ranking system orders them against a query using signals including topical relevance, link-based authority, user interaction patterns, freshness, and page experience. The system was never a fixed formula and has become less formula-like as machine-learned ranking components took over. What matters commercially is that ranking position determines click share steeply — the top result captures a large multiple of the fifth — and that the number of pixels above the first organic result kept growing.

Generative answer layers now sit above it. Google’s AI Overviews and AI Mode synthesise an answer and cite sources. The Pew Research Center study, based on 900 US adults from its KnowledgePanel and 68,879 unique Google searches during March 2025, found AI summaries appeared on 18% of searches, and that summary presence correlated with roughly halved click rates on traditional links. Clicks on links inside the summary itself occurred in 1% of visits. Sessions ended entirely after 26% of pages carrying a summary, against 16% without. Trigger rates scaled with query complexity: 8% of one- or two-word searches produced a summary, 53% of searches with ten or more words, and 60% of question-format searches.

That last detail carries a strategic implication that is easy to miss. The queries most likely to trigger a generative summary are precisely the long, question-shaped, informational queries that content marketing was built to capture. A decade of best practice told marketers to answer questions comprehensively. The surface that rewarded that behaviour is the surface being absorbed.

Social recommendation feeds are the third system. Reach here is allocated by predicted engagement rather than by follower relationship. A post from an account with 200 followers can outperform one from an account with 200,000, and the same account’s reach can vary by an order of magnitude between posts. The practical consequence is that organic social reach behaves like a lottery with a skill component rather than like an audience you own, which makes it poor for reliable reach delivery and good for asymmetric upside.

Video and long-form recommendation systems — YouTube most importantly — behave differently again, because watch-time maximisation rewards content that holds attention rather than content that provokes a quick reaction, and because the catalogue keeps working. A YouTube video can accumulate views for years, which makes it one of the few remaining places where the classical compounding-asset argument holds with force.

Community and third-party surfaces form the fifth. Reddit threads, industry forums, review platforms, Slack and Discord communities, podcast mentions, and comparison sites. Pew’s data noted Wikipedia, YouTube and Reddit together accounted for 15% of AI summary sources. Presence in these places is largely unbuyable and increasingly influential, both because humans consult them and because language models trained and grounded on the open web treat them as evidence.

The unifying property across all five is that you are a supplier to someone else’s ranking system, and the ranking system’s objective is its own retention and revenue, not your traffic. Organic content is not free distribution. It is distribution rented on terms that can be changed without notice and without compensation, paid for in production cost rather than media cost. Recognising that removes the moral framing — organic is not the virtuous choice — and returns the decision to arithmetic.

The 2026 cost curve for bought attention

The clearest way to see what happened to paid media pricing is to read the sellers’ accounts rather than the buyers’ complaints. Buyers report rising costs every year; that is a permanent feature of the profession. Sellers report volume and price separately, and the separation is where the information sits.

WARC’s December 2025 forecast put the global advertising market at $1.19 trillion for 2025 with 8.9% growth, upgraded by 1.5 percentage points from its September position, followed by 9.1% growth in 2026 and 7.9% in 2027. The same analysis noted that Alphabet, Amazon and Meta would collectively absorb the vast majority of incremental global ad spend between 2025 and 2027. A market growing at 9% while three companies capture most of the increment is a market where the price of access is set by an oligopoly with improving prediction models.

Meta’s own disclosure gives the cleanest read on unit price. Ad impressions across its family of apps rose 19% year over year in the first quarter of 2026, and average price per ad rose 12%. Revenue reached $56.31 billion, up 33% year over year, or 29% in constant currency. Alphabet reported $109.9 billion in consolidated first-quarter 2026 revenue, up 22%, with Google Search and other advertising at $60.4 billion, up 19%, and YouTube advertising at $9.88 billion, up 11%.

Read those figures against a typical advertiser’s gross margin trajectory. A consumer brand with 55% gross margins that saw no price increase of its own absorbed a 12% rise in unit media cost against flat contribution per order. If media cost was 20% of revenue, a 12% media inflation rate with static conversion rates removes roughly 2.4 percentage points of contribution margin in a single year. For businesses running at single-digit net margins, that is the difference between profitable growth and unprofitable growth, and it explains why paid-heavy DTC brands spent 2025 and 2026 rebuilding toward retention and owned channels rather than acquisition volume.

There is a counter-reading that deserves fair treatment. Rising prices in an auction are evidence that other advertisers are finding the inventory worth the price. If paid social returns had collapsed, bidders would have withdrawn and prices would have fallen. Sustained price inflation alongside sustained volume growth is, on its face, evidence that paid media works well enough for enough advertisers to keep bidding. The pessimistic interpretation — that advertisers are systematically fooled by platform-reported conversions — requires believing that a large population of sophisticated buyers with independent measurement is wrong for years at a time. Some are. Not all are.

The honest synthesis is that both readings are partially right, and which applies depends on the advertiser’s measurement quality. Companies running rigorous holdout tests generally find a subset of their paid spend is strongly incremental and another subset is not, and they cut the second. Companies relying on platform-reported return on ad spend generally overstate paid performance, and the overstatement grows as automated campaign types absorb more brand and retargeting traffic.

Paid and organic compared on the dimensions that drive allocation

The table describes tendencies, not guarantees, and every row has category exceptions. Its practical use is as a checklist: an allocation argument that ignores three or more of these rows is incomplete.

One more pricing dynamic is worth isolating. Google Network revenue declining 4% year over year while Search grew 19% describes a specific reallocation: money moving out of the open web and into owned surfaces. For publishers and content businesses, that is a direct revenue shock. For advertisers, it is a narrowing of the places where cheaper attention can be found, which raises the average price of the remaining supply. The concentration of ad spend into a small number of owned-and-operated environments is itself an inflationary force, independent of demand.

The organic traffic contraction, measured honestly

The most common error in discussing organic performance in 2026 is treating a click decline as a demand decline. They are different events with different commercial consequences, and conflating them produces bad strategy in both directions.

What the evidence supports firmly: when a generative summary appears above search results, the probability that a user clicks a traditional result falls substantially. Pew’s March 2025 dataset found 8% click rates with a summary present against 15% without, with summary presence on 18% of searches. Independent industry studies published through 2025 and 2026 reported click-through rate declines for top-ranking pages in similar ranges, with figures around 58% reduction circulating widely for high-position organic listings on summary-bearing queries. Industry-side figures of this kind come from panel or clickstream samples with real methodological limits, and should be read as directional rather than precise.

What the evidence also supports: total search demand did not fall. Alphabet’s leadership stated in the first-quarter 2026 earnings call that queries were at an all-time high and that AI Overviews and AI Mode were driving greater search usage and growth in overall queries. Both things can be true simultaneously, and their combination is the actual problem. More searches, distributed across more surfaces, with a lower click rate per search, produces an environment where a site can lose traffic while its category grows.

The distributional detail matters more than the average. Queries most affected are informational and question-shaped, because those are the queries generative summaries answer well and the queries where a user’s need is fully satisfied by a paragraph. Queries least affected are navigational, transactional, and those requiring a decision among specific options — someone comparing three products, checking stock, booking a slot, or reading reviews still needs to arrive somewhere. The contraction is concentrated precisely in the content type most companies produced most of: the explanatory top-of-funnel article.

This produces an uncomfortable conclusion for content teams. A large share of published content inventory was built to capture query intents that no longer generate visits. That inventory is not worthless — it may still be cited inside AI answers, still support topical authority, still convert the reduced traffic it receives — but its traffic yield per unit of production cost has fallen, in some categories severely. Treating that as a temporary dip rather than a structural repricing has led companies to keep funding the wrong content type.

The Content Marketing Institute and MarketingProfs annual B2B research, based on 1,015 B2B marketers surveyed between 24 June and 14 August 2025 and published in October 2025, showed the profession adapting unevenly. Fifty-nine percent rated their content marketing somewhat or highly effective, 31% reported neutral or mixed results, and 10% called their efforts ineffective. Content relevance and quality was the leading improvement driver at 65%, ahead of team skills at 53%, sales alignment at 45% and technology at 43%. Measuring effectiveness ranked third among challenges at 33%, behind creating content that prompts action at 40% and resource constraints at 39%. Owned media ranked third among 2026 investment priorities at 32%, behind AI tools at 45% and events at 33%, while human resources ranked last at 9%.

That ordering is a signal in itself: teams are buying tools and stages rather than the people who make content distinctive, at exactly the moment when distinctiveness became the scarce input.

There is a second contraction that gets less attention and hurts more slowly. Organic social reach continued its long decline as feeds shifted further toward recommended content from accounts users do not follow. The practical effect is that an audience you spent years building on a social platform now delivers a fraction of the reach it once did, and the platform will sell you back access to that same audience. This is the clearest case in marketing where an apparently organic asset was always a leased one.

None of this justifies abandoning organic work. It justifies changing what organic work means: fewer explanatory articles competing with a summarising machine, more content with information the machine cannot generate — proprietary data, original testing, named expert opinion, customer evidence, tooling, community. That shift is defensible on the evidence, and it is also considerably harder and more expensive than what it replaces.

Answer engines rewrote the discovery layer

A third distribution surface now sits alongside paid and organic, and it does not behave like either. Assistants and answer engines — Google’s AI Mode, ChatGPT, Gemini, Perplexity, Copilot, and the assistant layers embedded in browsers, operating systems and retail apps — mediate an increasing share of commercial research. The strategic problem is that this surface currently offers organic-style unbuyability with paid-style commercial intent behind it.

The monetisation direction is not ambiguous. Google’s chief business officer Philipp Schindler told investors in the first-quarter 2026 earnings call that Gemini had “significantly expanded our ability to deliver ads on longer, more complex searches that were previously really difficult to monetise,” and described agentic shopping experiences as additive, saying they would “really transform how we shop from discovery to decisions.” Search Engine Land’s March 2026 analysis of AI Mode monetisation argued the rollout would be deliberately gradual, with Google monetising lightly to build data and grind down competitors in assistant conversations rather than maximising near-term revenue. OpenAI moved through 2026 toward advertising inside ChatGPT, with reporting through the year describing phased rollouts across markets and response surfaces.

For a marketer, three consequences follow, and they cut in different directions.

The first is that a growing share of commercially useful exposure produces no session and no analytics event. A prospect who asks an assistant to compare three vendors, receives a synthesised answer citing your documentation, and forms a preference has been influenced without visiting your site. Conventional measurement records nothing. The company then reduces investment in the content that produced the influence, because the dashboard shows it delivering less. This is a measurement-induced strategy error, and it is the single most likely way well-run companies will misallocate budget in 2026 and 2027.

The second consequence is more encouraging and has stronger evidence than most of the surrounding commentary. Where assistant referrals do produce a visit, those visits tend to convert at higher rates than classical organic visits. Multiple independent 2025 and 2026 analyses of site-level data reported conversion-rate multiples for assistant-referred traffic against organic search traffic, with figures commonly in the range of four to five times, and some datasets reporting absolute conversion rates for assistant referrals in the high single digits to high teens. Treat the precise multiples with caution — sample sizes vary, self-selection is severe, and the traffic volumes are small relative to organic — but the direction is consistent enough across independent datasets to be taken seriously: assistant referrals arrive later in the decision process, pre-qualified by a conversation that has already narrowed the option set.

The third consequence is competitive rather than technical. Citation inside an answer is winner-take-most in a way that ranking never was. A search results page shows ten links; an assistant answer typically names two to five options. Falling from position eight to position twelve on a results page costs a small amount of traffic. Falling out of an answer’s shortlist costs everything on that query. Answer-layer visibility has a steeper cliff than search ranking, which makes it a higher-variance investment.

What actually drives inclusion in these answers is less mysterious than the emerging consultancy market implies, and less controllable than it promises. Models ground their answers in retrieved documents, and retrieval favours sources that are well-structured, clearly attributed, frequently referenced elsewhere, and unambiguous about entities. Third-party corroboration matters heavily — being named in independent comparisons, review platforms, and community discussion appears to influence inclusion more reliably than anything on a company’s own domain. Pew’s finding that Wikipedia, YouTube and Reddit together supplied 15% of AI summary sources supports that reading.

The honest position on the whole surface is that it is strategically important, currently unbuyable at scale, poorly measurable, and about to become partially buyable. Companies should be building presence in it now, funding that work from the content budget rather than expecting the ad budget to solve it, and instrumenting whatever measurement they can — server-log analysis of assistant crawlers, referral tracking for the visits that do occur, and periodic manual prompt testing across the major assistants to establish whether the brand appears at all for its core commercial questions.

That last practice, running a fixed set of buying-intent prompts across assistants on a monthly schedule and recording whether and how the brand is described, is cheap, takes an afternoon, and gives a company more useful information about its answer-layer position than any tool currently on the market. Very few companies do it.

Attribution is the real reason this argument never ends

Strip away the professional politics and the paid-versus-organic debate reduces to a measurement dispute. The two channels are measured by systems with opposite biases, and nobody reconciles them.

Paid media is measured by the seller. Ad platforms observe an impression or click, then observe a conversion, then claim credit according to their own attribution window and model. Every incentive in that arrangement points toward over-crediting. Not through fraud, but through structure: a platform that sees only its own touchpoints will attribute conversions it merely witnessed. View-through windows, cross-device graphs, and modelled conversions all expand the set of outcomes a platform can plausibly claim. Sum the reported conversions across a mid-sized advertiser’s platforms and the total routinely exceeds actual orders by a wide margin.

Organic content is measured by last-click analytics, which has the opposite bias. A blog post read in January, remembered in March, and followed in June by a direct visit and purchase produces an analytics record showing direct traffic. The content that created the demand receives no credit at all. Last-click measurement systematically transfers credit from demand creation to demand capture, which in practice means from content and brand work to paid search and retargeting.

The consequence is a reinforcing loop that operates in almost every company that has not deliberately broken it. Paid appears productive because its measurement over-credits. Organic appears unproductive because its measurement under-credits. Budget moves from organic to paid. Organic demand declines with a lag of six to eighteen months. Branded search volume falls. Paid costs rise because the paid campaigns were harvesting the demand organic created. The company concludes that paid media is inflating and asks for a bigger budget. This loop is not a hypothetical; it is the standard failure mode of performance-led marketing organisations, and it takes two to three years to become visible.

Several developments made the measurement problem worse rather than better. Consent requirements under GDPR and the ePrivacy Directive reduced the share of European sessions that can be tracked at all, with consent rates varying widely by country and vertical. Browser restrictions shortened cookie lifetimes. Apple’s App Tracking Transparency cut mobile signal for a large share of iOS users. Server-side tagging and conversion APIs restored some signal, at the cost of pushing measurement into infrastructure that most marketing teams cannot audit. Google’s decision to retain third-party cookies in Chrome and wind down the Privacy Sandbox APIs removed a deadline but did not restore the underlying signal quality, because the erosion had come mostly from other browsers, consent frameworks and platform policy.

Platforms responded by modelling. Modelled conversions fill observation gaps with statistical estimates. This is legitimate methodology and it is also unverifiable from the buyer’s side. When a platform reports that a campaign produced 400 conversions and a sizeable share of those are modelled rather than observed, the advertiser is being asked to accept the seller’s estimate of the seller’s own performance. No other category of business purchase works this way.

The practical response is not to distrust everything, which paralyses decisions, but to establish a hierarchy of evidence. Randomised or geo-based experiments sit at the top, because they measure counterfactuals. Marketing mix modelling sits second, because it uses aggregate time-series data and is immune to individual-level tracking loss, though vulnerable to specification error and collinearity. Platform-reported attribution sits third, useful for within-platform tuning decisions and unreliable for cross-channel allocation. Last-click analytics sits fourth, useful for operational diagnosis and misleading for credit assignment.

Any company arguing about paid versus organic without at least one source of tier-one or tier-two evidence is not having an analytical discussion. It is having a political one. The distinguishing test is simple: ask what result would change the participants’ minds. If no answer exists, the conversation is about territory.

Incrementality tests and what they keep revealing

Experiments are the only method that answers the allocation question directly, and the reason they are underused is not ignorance but cost: they require deliberately withholding marketing from some customers, which feels like burning money and is politically difficult to authorise.

The two workable designs are conversion lift tests and geographic experiments. Conversion lift tests hold out a randomised share of the target audience from seeing a campaign and compare conversion rates between exposed and held-out groups. Platforms offer these natively, which introduces the obvious concern that the seller designs the test, though the randomisation itself is usually sound and the results are frequently unflattering to the seller. Geographic experiments switch spend off in a matched set of regions and compare outcomes against control regions, using synthetic control or difference-in-differences methods. Geo tests measure total channel effect including offline and untracked conversions, which is their main advantage over platform lift tests.

Google’s investment here is notable, because it comes from a company whose reported attribution benefits from being unaudited. At Google Marketing Live in May 2026 the company announced Meridian GeoX, a geographic incrementality testing tool with an open-source codebase intended to provide what Google described as ground-truth validation of media channel performance, with testing beginning later in 2026, alongside Meridian Studio, an enterprise platform on Google Cloud for managing high-volume marketing mix models. Google’s Gaurav Bhaya framed the direction as “measurement is your engine for growth in the AI era.”

Interpreting a platform building better tools for measuring its own incrementality requires some care. The charitable reading, and probably the correct one, is that Google’s largest advertisers now demand experimental evidence as a condition of budget growth, and supplying the tooling is cheaper than losing the argument. The strategic reading is that a company with genuinely incremental inventory benefits from a market where incrementality is measurable, because it wins that comparison against competitors whose inventory is less incremental.

What the accumulated body of published and practitioner experiment results consistently shows, across categories, is a pattern worth stating carefully because it is the most decision-relevant finding in the whole debate:

Branded search advertising is frequently far less incremental than platform attribution suggests. When companies pause bidding on their own brand terms in test regions, total conversions often fall by much less than the attributed conversions the paid campaigns were reporting, because organic listings capture most of the traffic. The incremental share varies with competitive brand bidding, category, and whether the organic listing occupies the first position, so the correct response is to measure rather than assume, but the direction of the bias is well established.

Retargeting is the second most commonly overstated tactic, for the same structural reason: it addresses people who have already demonstrated intent.

Prospecting and broad-reach paid campaigns are typically less flattered by attribution and more incremental than they appear, because their effects are diffuse, delayed and partly captured by other channels. Companies that cut prospecting because its reported return on ad spend looked poor have repeatedly found total revenue falling more than the cut implied.

Organic content and brand activity usually measure better under experimental and modelling approaches than under last-click. Marketing mix models frequently assign content, SEO and brand media a larger share of contribution than digital analytics does, though model specification choices influence this and honest practitioners report wide confidence intervals.

The operational lesson from all of this is that the paid-versus-organic split is the wrong unit of analysis. The useful unit is the individual tactic, tested. A media plan is not one paid decision and one organic decision; it is thirty decisions, of which perhaps eight are strongly incremental, twelve are modestly incremental, and ten are near-zero. Finding which is which is worth far more than winning an argument about categories.

A practical caution on running these tests: they need enough scale to detect an effect. A business with 40 conversions a month cannot run a two-week holdout and learn anything, because the noise exceeds any plausible signal. Small advertisers should rely on longer before-and-after comparisons with careful attention to seasonality, on channel-level pauses of at least six to eight weeks, and on tracking branded search volume and direct traffic as leading indicators of demand-creation health.

Marketing mix modelling as the referee

Marketing mix modelling returned to prominence for an unglamorous reason: it does not need user-level tracking. When individual identifiers degraded, a method built on aggregate weekly spend, sales, price, distribution, seasonality and external factors became attractive again.

The method regresses an outcome — sales, orders, leads — against marketing inputs and controls, then estimates each input’s contribution, usually with diminishing-returns curves and adstock terms that model delayed effects. Because it works on aggregates, it captures effects that tracking cannot see: offline conversions, cross-device journeys, assistant-mediated influence, word of mouth partly correlated with media, and brand effects with long lags. That makes it the only widely available method that gives organic content and brand media a fair hearing alongside paid performance channels.

Its weaknesses are real and should be stated by anyone recommending it. Collinearity is the biggest: when paid search and content investment move together over time, the model struggles to separate their contributions, and small specification changes can shift credit substantially between them. Data requirements are demanding — typically two to three years of weekly data, with genuine variation in spend levels. Models are sensitive to which controls are included, and omitting a variable that drove sales assigns its effect to whatever correlates with it. And the outputs are estimates with confidence intervals that vendors sometimes present with more precision than the underlying data supports.

Open-source availability changed the accessibility picture. Google’s Meridian and Meta’s Robyn both made Bayesian and automated modelling approaches available without licence fees, which pulled the technique down-market from large-advertiser territory into reach of mid-sized companies with analytical capability. The bottleneck is no longer software cost; it is data hygiene and the availability of someone who can specify a model and interpret it honestly.

The most productive way to use modelling in the paid-versus-organic question is not to accept its point estimates as truth but to use it as a hypothesis generator that experiments then test. A model suggesting content contributes 18% of conversions with a wide interval is not proof. It is a reason to run a geo test on content-driven activity, or to examine what happened in a period when content output paused. Modelling proposes, experiments confirm, and platform attribution executes within whatever envelope the first two establish. That hierarchy is the whole discipline of modern measurement, compressed.

Google’s Meridian Studio announcement in May 2026 reflects where the market moved: from modelling as an annual consulting exercise to modelling as a running system refreshed continuously and connected to bidding decisions. That direction has an obvious risk attached. A model refreshed weekly and connected to automated budget allocation can propagate specification error at speed. The safeguard is calibration against experiments, which is precisely why Meridian GeoX and Meridian Studio were announced together.

One practical warning for mid-sized companies. Modelling rewards spend variation, and most companies deliberately avoid variation because it looks like poor planning. If paid budget has been flat at €40,000 a month for two years, no model can estimate its diminishing-returns curve, because the data contains no information about what happens at €20,000 or €70,000. Deliberately varying spend — stepping a channel up and down over quarters, staggering regional launches, pausing tactics on a rotation — has measurement value that exceeds its short-term efficiency cost. Very few marketing plans are built to generate information about themselves.

The compounding asset argument, tested against evidence

The strongest case for organic content has always been that it builds an asset. Pay once, receive traffic indefinitely, and watch cost per visit fall toward zero as the denominator grows. The case is sound in principle and has become materially weaker in practice, and the reasons deserve careful examination rather than either dismissal or nostalgia.

Consider the arithmetic that made the argument compelling. An article costing €500 to produce that attracts 500 visits a month for three years delivers 18,000 visits at €0.028 per visit. The same 18,000 visits bought at €1.20 per click cost €21,600. The ratio is so extreme that it dominated content marketing advocacy for a decade, and where it still holds, it remains the best deal in marketing.

Three things degraded that arithmetic. Content decay accelerated. Published pages lose traffic over time as competitors publish, as search intent shifts, and as ranking systems favour freshness in more categories. Maintaining a library of 400 articles at competitive quality is a permanent operating cost that the original arithmetic ignored, and it typically consumes 20% to 40% of a content team’s capacity once the library reaches that size. Competition raised the production cost floor. The €500 article that ranked in 2016 does not rank in 2026, because the results page it competes on contains original research, expert commentary, proprietary data and structured tooling. Producing something competitive now costs multiples of that figure. And the click yield per ranking position fell, for the reasons the Pew data documents.

Combine the three and the compounding argument does not vanish, but its terms change fundamentally: higher production cost, shorter useful life, lower traffic per position, and a longer period before payback. A content programme that once paid back in nine months may now pay back in twenty-four, which changes it from a marketing tactic into a capital allocation decision requiring genuine patience and a stable strategy.

There is a version of the compounding argument that has strengthened rather than weakened, and it is where the serious money now sits. Content that cannot be synthesised from existing web text — original research with proprietary data, tools and calculators, product documentation, community, video with a recognisable presenter, and named expert opinion with a track record — has become more defensible precisely because generative systems can trivially reproduce everything else. Scarcity moved from the ability to publish to the ability to publish something that could not have been generated. A benchmark study nobody else can run, an index built from your own transaction data, a tool that performs a calculation, a video series with a face and a voice: these accumulate authority, get cited by assistants, get linked by publishers, and cannot be commoditised by the next model release.

The corollary is that the compounding case now supports a smaller volume of much more expensive content. That inverts a decade of content operations built around throughput. Teams organised to publish twelve articles a month are structurally wrong for an environment that rewards two pieces of original work a quarter, and the reorganisation is painful because it changes who is needed.

A second form of compounding deserves equal weight and gets far less attention: owned audience. An email list, a subscribed newsletter, an app install base, a community, a customer database. These are the only marketing assets that are genuinely owned rather than leased, they are unaffected by ranking changes and auction inflation, their marginal distribution cost is near zero, and their performance is measurable. Every euro of paid or organic acquisition that ends without capturing a means of contacting the person again has purchased a single event rather than a relationship.

The Gartner finding that loyalty and retention spending fell 29% since 2024 and now sits under 15% of total media spend, while awareness and conversion together consume 62.6%, describes a market moving in the opposite direction from where the compounding logic points. That is either a collective error or a rational response to short-term pressure. The interpretive judgment here — and it is a judgment, not a measurement — is that it is mostly the former, driven by measurement systems that report acquisition well and retention badly, and by executive time horizons shorter than retention payback periods.

The speed premium and where it is worth paying

Paid media’s defining advantage is not reach or targeting. It is latency. A campaign can be live in an afternoon, at chosen volume, in chosen markets, with chosen creative, and can be stopped as quickly. Nothing in organic distribution offers control over timing, and there are commercial situations where timing is the entire value.

Product launches are the clearest case. A launch has a window in which coverage, retail attention, and category interest coincide. Organic content published on launch day reaches its distribution peak months later, by which time the window has closed. Paying for reach in the launch window is not inefficiency; it is buying the only distribution available at the moment it matters.

Seasonal peaks work identically. A retailer with 40% of annual revenue in six weeks cannot wait for content to mature into the peak. Insurance renewal cycles, tax deadlines, academic admissions calendars, wedding season, and travel booking windows all concentrate demand into periods where the ability to appear on demand has value that exceeds its unit cost.

Testing is the underrated case and probably the highest-return use of paid budget in the whole plan. Paid media is the cheapest available instrument for learning what messages, offers, audiences and price points work, because it returns statistically usable results in days rather than quarters. A company can test six positioning statements across a defined audience for a few thousand euros and get a clear ranking within a week. The same learning through organic content takes two quarters and returns confounded results. Companies that treat a portion of paid budget explicitly as research spend, with learning rather than immediate return as the success criterion, generally get better organic content, better landing pages and better sales messaging as a byproduct.

Geographic and segment expansion is a fourth case. Entering a new market with no brand presence, no backlinks, no local content and no community means organic distribution starts from zero and takes years. Paid media provides immediate presence while the organic base is built. Attempting market entry organically to save money usually costs more, because it extends the period of paying fixed costs with no revenue.

Countercyclical opportunity is a fifth, and the most commercially interesting. Auction prices fall when competitors withdraw. When a category’s advertisers cut budgets in a downturn, the cost of the same attention drops, and companies with cash can buy reach at prices unavailable in normal conditions. This requires the financial ability to spend against short-term pressure, which is exactly what most companies lack at precisely the moment it becomes available.

Against all of that, the speed premium should be recognised as a premium — a real cost paid for real convenience. The mistake is not paying it. The mistake is paying it permanently for activity that has no timing requirement. Evergreen category education has no deadline. Ongoing awareness in a category with steady demand has no deadline. Paying auction prices year after year to distribute content that could have been distributed organically is the specific behaviour that produces the margin compression described earlier.

A useful discipline: for each line in the paid plan, state the timing reason. Launch window, seasonal peak, test, market entry, competitive response, or price opportunity are all valid reasons. “We have always run this” is not a timing reason, and lines that cannot supply one are candidates for a holdout test.

The quiet failure modes of organic content

Advocacy for organic content tends to skip its failure modes, which is unhelpful because they are predictable and mostly avoidable. Six recur across companies of every size.

The volume trap. Teams measure output because output is countable, then publish quantity that no longer earns distribution. The failure is structural rather than lazy: a content calendar creates an obligation to publish weekly regardless of whether anything worth publishing exists that week. The result is a library of adequate pages, none distinctive, competing against results pages that reward distinctiveness. A company publishing forty mediocre articles a year would in most categories be better served by four genuinely original ones and a mechanism for distributing them.

Confusing production with distribution. Content teams are usually staffed for creation and rarely staffed for promotion. Publishing is the midpoint of the work, not the end. A piece of original research needs outreach to journalists and analysts, syndication, a presentation version, a video version, a social sequence, an email sequence, sales enablement, and follow-up over months. Companies that spend €20,000 producing research and nothing distributing it have wasted most of the €20,000. The distribution-to-production budget ratio in successful content programmes is frequently 1:1 or higher, and in most companies it is closer to 1:10.

Undefined commercial intent. Content briefs specify topic and keyword and omit what the reader should be able to do afterwards, which stage of the buying process the piece serves, and what the company gains. The result is traffic that does not convert, followed by a conclusion that content does not work. The Content Marketing Institute research found creating content that prompts action was the leading challenge at 40%, ahead of resource constraints, which suggests this is the modal failure rather than an unusual one.

Ignoring maintenance. Every published page is a liability as well as an asset. It ages, its claims go stale, its links break, its screenshots become wrong, its rankings slip. A library of 300 pages requires a systematic refresh cycle. Companies that keep publishing without a decommissioning and refresh process end up with a site where a large share of pages produce nothing and dilute the quality signal of the rest.

Attribution starvation. Because organic content measures badly, it loses budget battles it should win, and then loses them again because the reduced investment produces reduced results, which confirms the original judgment. Breaking that loop requires deliberately instrumenting content with intermediate metrics that do respond within a quarter — assisted conversions, branded search volume, direct traffic, email list growth, sales-cited content, deal-stage acceleration, and demo requests from known content consumers — and reporting those alongside revenue rather than instead of it.

Timeline dishonesty at the point of approval. Content programmes are frequently approved on the basis of six-month expectations and evaluated on that basis, when the realistic payback in a competitive category in 2026 is twelve to twenty-four months. The programme is then cancelled at month seven, having produced most of its cost and little of its return. The failure occurred in the business case, not in the execution. Approving a content programme without agreeing an evaluation horizon that matches its physics is a decision to waste the money.

A seventh failure mode deserves separate mention because it became common only recently: building an organic strategy around query intents that generative summaries now absorb. Any content plan still oriented around comprehensive answers to short informational questions is investing in a surface that Pew’s data shows produces half the clicks it did, on the query types most likely to trigger a summary. Auditing an existing content library by query type — informational and question-shaped versus comparative, transactional, and decision-support — is one of the highest-value analyses a content team can run in 2026, and it usually reveals that the majority of historical investment sits in the exposed category.

Paid media’s blind spots, itemised

The failure modes on the paid side are better documented in practitioner literature and, oddly, less often acted on, because acting on them requires reducing spend on activity that reports well.

Buying demand you already had. The single largest source of wasted paid budget in most accounts is spend directed at people who were already going to convert. Brand-term bidding, site retargeting, existing-customer targeting inside broad automated campaigns, and email-list lookalikes that substantially overlap the actual list. All of it reports excellent returns and much of it is not incremental. The remedy is experimental measurement, and the reason it is rare is that it usually requires a performance team to prove a portion of its own reported results are illusory.

Automation drift. Automated campaign types redistribute spend toward whatever satisfies the objective most cheaply, which frequently means existing customers and existing intent. Without explicit exclusions, new-customer acquisition targets, and separate reporting for new versus returning buyers, an automated campaign will gradually convert itself into a retention campaign priced as acquisition. The dashboard will look better as this happens, because retention converts more cheaply than acquisition.

Creative fatigue treated as an audience problem. When performance declines, the standard response is to change targeting or raise bids. In consumer categories the actual cause is usually that the audience has seen the creative too many times. Frequency and unique-reach diagnostics are available and routinely ignored, and the cost of ignoring them is paying inflated prices for impressions with declining response.

Landing-page and offer neglect. Paid media traffic hits the same pages as everything else, and the pages are built for nobody in particular. Doubling landing page conversion rate halves cost per acquisition without touching the media plan, and it is cheaper than any bidding change. Companies spending €50,000 a month on media and nothing on conversion-rate work are making an arithmetic error, not a strategic one.

Measurement inflation across platforms. Each platform claims conversions it saw. Summed, the claims exceed reality. Companies that report channel performance from platform dashboards without reconciliation against actual orders build plans on numbers that do not add up, and then allocate budget toward whichever platform is most aggressive in its attribution windows. The reconciliation discipline is simple — total platform-reported conversions divided by actual new customers gives an inflation ratio — and its absence is near-universal in mid-market companies.

Structural margin blindness. Paid campaigns are evaluated on return on ad spend, which ignores gross margin, returns, discount depth, and the cost to serve. A 4:1 return on ad spend is excellent at 70% gross margin and loss-making at 25% with a 30% return rate. This is not a subtle point and it is missed constantly, particularly in fashion and consumer electronics.

Dependency accumulation. A company whose revenue depends on two auction-based channels has outsourced its demand to suppliers who set their own prices, change their own rules, and compete with their customers for margin. Meta raising average price per ad by 12% year over year is a supplier exercising pricing power over a customer base with limited alternatives. The strategic risk is not that any individual campaign fails; it is that the entire revenue base has a cost input that a third party controls and that has risen every year for a decade.

The pattern uniting all seven is that paid media’s problems are mostly invisible on the reports paid media generates. That is the mirror image of organic content’s problem, whose value is invisible on the reports it generates. Both distortions push in the same direction, which is why the correction almost always involves moving some budget from paid to owned and organic work — and why that correction is almost always resisted by whoever holds the paid budget.

The 60/40 evidence and its limits in 2026

The most cited empirical answer to the allocation question comes from Les Binet and Peter Field, whose analysis of IPA Effectiveness Awards case studies proposed that a roughly 60/40 split between long-term brand building and short-term activation produced the strongest business results across categories. The finding has been extended and refined across multiple publications and remains the closest thing marketing has to a general allocation heuristic derived from a large body of cases.

The underlying logic is worth stating precisely because it is frequently misrepresented. Brand building works slowly, at low measured efficiency, and creates demand and price tolerance that persist. Activation works quickly, at high measured efficiency, and converts demand that already exists. Because activation always measures better than brand building, an organisation that allocates by measured efficiency will systematically underinvest in brand building, and the underinvestment only becomes visible after the accumulated brand effect has depleted. The 60/40 proposal is a corrective against that measurement bias rather than a claim about optimal ratios in any specific case.

Binet himself has been clear that it is not a rule. In interviews he has stressed that the split varies by category, brand size, market growth, purchase cycle length, and whether the brand is in a challenger or defender position. Ehrenberg-Bass Institute research, associated with Byron Sharp’s work on how brands grow, supports a complementary conclusion from a different angle: growth comes predominantly from increasing penetration among light and non-buyers rather than from deepening loyalty among heavy buyers, and reaching those light buyers requires broad, continuous, distinctive advertising rather than precisely targeted activation.

Applying either framework to the 2026 paid-versus-organic question requires a translation step that is often botched. Brand building is not synonymous with organic content, and activation is not synonymous with paid media. Broad-reach paid campaigns, connected TV, sponsorships and creator partnerships are paid brand building. Paid search on high-intent commercial queries is paid activation. A comparison article ranking for “best CRM for small business” is organic activation. A research report that shapes how a category thinks about a problem is organic brand building. The 60/40 axis and the paid/organic axis are perpendicular, and plotting a media plan on both grids simultaneously is more informative than either alone.

Done properly, that two-axis view produces four quadrants, and most companies find their spending concentrated in one of them. Paid activation dominates the plans of performance-led ecommerce and lead-generation businesses. Organic activation dominates the plans of SEO-led affiliate and comparison businesses. Paid brand building dominates large consumer advertisers. Organic brand building — research, thought leadership, community, video presence — is the quadrant most companies fund least and the one where the answer-layer shift has most increased the return, because it produces exactly the material that assistants cite and competitors cannot generate.

There are three specific reasons to treat the 60/40 figure with more caution in 2026 than in 2018, and stating them fairly matters more than defending the heuristic.

First, the case base skews toward large consumer brands with substantial budgets, and the ratio’s applicability to a company spending €8,000 a month is unestablished. Below a certain absolute spend, broad-reach brand building may deliver too little reach to produce any measurable effect, making concentration on activation the rational choice until scale permits otherwise.

Second, the cost of brand building rose relative to its historical level because attention fragmented across more surfaces and the price of reach on the major platforms increased. The same 60% of budget buys less brand effect than it did, which shifts the arithmetic optimum somewhat toward activation, though not by an amount anyone has credibly quantified.

Third, the measurement asymmetry the heuristic corrects for has grown worse, not better. Under-crediting of brand and content activity increased as tracking degraded and as assistant-mediated influence became untrackable. That argues for the correction being larger, not smaller. The two forces work against each other and their net effect is genuinely unknown.

The defensible use of the framework is therefore not to adopt 60/40 as a target but to use it as a floor test: if long-term demand-creation work receives less than about 30% of total marketing investment, the plan is very likely under-invested in it, for reasons the measurement system will never reveal. That is a weaker claim than the original and it survives scrutiny.

Ecommerce and DTC economics under paid pressure

Direct-to-consumer ecommerce is the category where the paid-versus-organic question has the sharpest financial consequences, because the business model has almost no tolerance for acquisition cost inflation.

The arithmetic is unforgiving. A brand with 60% gross margin and a €70 average order value earns €42 contribution per first order before fulfilment, payment processing and returns. Subtract €8 fulfilment, €2 processing and a 12% return rate and contribution falls to roughly €27. If blended customer acquisition cost is €35, every first order loses money and profitability depends entirely on repeat purchase. A 12% annual increase in paid media unit costs, unmatched by price increases, moves that business from marginal to structurally unprofitable within two years.

That is the mechanism behind the DTC contraction that began in 2022 and continued through 2026. Businesses built during a period of cheap paid social distribution discovered their model had been subsidised by an auction inefficiency. The survivors did four things, and the list is a reasonable template for anyone in the category.

They raised average order value and repeat rate rather than chasing acquisition volume, through bundling, subscription formats, replenishment programmes and post-purchase merchandising. Contribution per customer is the only lever that makes rising acquisition cost survivable.

They built owned channels aggressively, particularly email and SMS, treating list growth as a primary metric rather than a byproduct. A brand that can generate 25% of revenue from owned channels at near-zero marginal distribution cost has a fundamentally different cost structure from one generating 5%.

They moved a share of content investment from articles to product-adjacent and creator content, because the informational article had weak commercial function in the category anyway and because short video content carries reach on the same platforms where paid distribution is bought, which makes the organic and paid programmes mutually reinforcing rather than separate.

They instrumented incrementality, most commonly through geo holdouts and periodic channel pauses, and typically found that a portion of their spend on retargeting and brand terms was not generating incremental revenue.

Where organic content genuinely works in this category is narrower than content advocates suggest and broader than performance marketers assume. Product-comparison and buying-guide content on high-intent commercial queries converts, survives generative summarisation reasonably well because the queries are decision-shaped rather than informational, and captures demand at a fraction of paid cost. Content built around the use of the product — technique, styling, maintenance, recipes, fitment — supports retention and repeat purchase, which is where the margin actually is. Category education content aimed at people who do not yet know the product exists is where most of the wasted content budget goes.

Retail media adds a complication specific to this sector. Brands selling through marketplaces and retailers now face a channel that sits closest to purchase, is priced at auction, and is in practice mandatory for shelf visibility. This functions as a distribution tax rather than a marketing choice, and it competes for the same budget as everything else while offering the least strategic optionality. Budgeting it as cost of sale rather than as marketing investment produces clearer decisions about what remains available for demand creation.

B2B SaaS and the long sales cycle problem

B2B software inverts several of the assumptions that make the ecommerce calculation tractable, and the inversions change the answer materially.

Deal values are high, sales cycles run from weeks to quarters, buying committees involve five to twelve people, and the purchase decision is influenced across a period during which most of the influence is untrackable. A €40,000 annual contract can justify a customer acquisition cost of €12,000 or more, which makes expensive paid clicks affordable in a way they never are in consumer categories. High-intent B2B search terms clearing at €25 to €80 per click are rational purchases when the conversion path is sound, and irrational when it is not, and the difference is usually the quality of the offer rather than the bidding.

The structural problem is that only a small share of a category’s potential buyers are in an active buying process at any moment. Research associated with the LinkedIn B2B Institute has popularised the estimate that roughly 5% of a business market is in-market at a given time, with the remainder forming impressions that will influence a purchase decision at some unknown future point. Whether the figure is exactly 5% matters less than its implication: paid search captures the small in-market fraction, and the much larger out-of-market population is reachable only through activity that builds recognition and preference before the buying process starts. That activity is predominantly content, community, video, events and broad-reach media — and its return is measurable only years later, or through modelling.

This is why B2B SaaS shows the widest gap between last-click attribution and experimental or modelled results. A prospect encounters a company’s research in month one, follows a founder’s commentary through months two to eight, attends a webinar in month nine, then searches the brand name in month eleven and converts through a paid brand-term click. Attribution credits the paid click. The paid click did nothing except charge the company for a customer its content had already won. Companies that pause brand-term bidding in B2B frequently find total pipeline barely moves.

The generative-answer shift affects this sector distinctively and, on balance, unfavourably in the short term and favourably in the medium term. Unfavourably because a large share of B2B content investment went into explanatory articles about categories, problems and methods, and those are exactly the queries generative summaries absorb. Favourably because assistants are becoming a serious vendor-shortlisting mechanism in software evaluation, and the sources they cite are documentation, comparison pages, review platforms, and community discussion. A company whose product documentation is public, precise and well-structured, whose presence on review platforms is substantial, and which is discussed by name in relevant communities, is well positioned in that mechanism. A company whose visibility rests on gated content and a marketing blog is not.

The gating question deserves a direct verdict. Gating content behind a form was rational when the resulting email address had value and the content was otherwise undiscoverable. In 2026 it has two costs that usually exceed its benefit: the content cannot be indexed, cited by assistants, or shared, and the leads it produces are frequently people who wanted the document rather than the product. Publishing openly and capturing intent through product trials, tools, calculators and self-service pricing generally produces fewer contacts and more pipeline. This is interpretation informed by practice rather than a measured finding, and it is contested by teams whose targets are set in marketing-qualified leads.

The practical allocation guidance for a B2B software company with a functioning product and a defined market: fund paid search on genuinely high-intent commercial and competitor queries, because the intent is real and the deal values justify the cost; fund paid social and connected TV for reach against the out-of-market majority only once there is enough budget for adequate frequency, since thin reach delivers nothing; fund content heavily but concentrate it in original research, product documentation, comparison material, customer evidence and a named human presence rather than category explainers; treat review platforms and communities as distribution channels requiring active investment rather than reputation management afterthoughts; and measure with holdouts on brand terms and retargeting first, because those are where the overstatement concentrates.

Local services and physical retail arithmetic

Local businesses face the simplest version of the question and receive the worst advice about it, largely because the marketing content industry writes for companies with national ambitions.

For a plumber, dentist, law firm, restaurant, gym or garage, demand is geographically bounded and largely non-expandable. The addressable market is the number of people within a travel radius who need the service, and no amount of content creates additional demand for emergency drain clearance. In a fixed-demand local market, marketing is a competition for share of existing intent rather than an exercise in demand creation, which shifts the optimum sharply toward capture channels.

The capture channels for local businesses are a specific and short list. A well-maintained Google Business Profile with accurate categories, hours, service areas, photographs and a steady flow of recent reviews. Presence in map results, which is driven by proximity, relevance and prominence, with reviews carrying substantial weight. Paid search on service-plus-location queries, which converts at rates that make relatively high costs per click viable when the average job value is in the hundreds. Local service ad formats where available. And local directory and aggregator presence, which continues to matter because those aggregators rank and because assistants cite them.

Content marketing for local businesses is worth doing in a form that bears little resemblance to standard content advice. Service pages for each service and each served area, written with genuine local specificity rather than templated town-name substitution. Answers to the questions customers actually ask on the phone, which converts well because the reader is close to booking. Photographic and video evidence of completed work, which functions as proof rather than as traffic acquisition. Reviews, which are the highest-return organic asset a local business has and which most local businesses collect passively. A systematic review-request process at job completion typically produces more commercial return than any content programme a business of that size could fund.

Physical retail sits between local services and ecommerce, with an additional measurement problem: much of the effect of digital marketing appears as footfall and in-store purchase, which digital attribution cannot see at all. This makes paid-versus-organic comparisons based on online conversion data actively misleading for retailers. A retailer measuring digital marketing on online conversions is measuring perhaps a quarter of the effect and allocating the whole budget on that quarter. Store-visit measurement, geo experiments comparing regional trading performance, and matched-panel work are the only methods that give an honest read, and geo experiments are within reach of any retailer with more than about fifteen locations.

The clearest allocation error in both categories is spending on brand-awareness activity aimed at a national audience when the business can only serve a radius. The second clearest is neglecting the review and profile work that costs almost nothing and drives most of the discovery. For most local businesses, the correct answer to paid versus organic is: fix the profile and reviews first, buy high-intent local search second, and treat everything else as discretionary.

Publishers, affiliates and the traffic collapse

For businesses whose revenue is a function of pageviews, the 2026 environment is not a strategic question but an existential one, and the paid-versus-organic framing barely applies because paid acquisition of traffic that monetises at display rates has never worked arithmetically.

The numbers describe the squeeze from both ends. Google Network revenue — the segment representing ads Google places on third-party sites — fell 4% year over year in the first quarter of 2026 to $6.97 billion, while Google’s own Search advertising grew 19%. The share of Google’s advertising revenue coming from the open web dropped below 9% from around 11% a year earlier. Money is leaving the independent web while total advertising grows at 9% annually. Publishers are losing share of a growing market, which is the worst possible position.

Simultaneously, the traffic that generates those declining rates is contracting. Pew’s data on halved click rates when summaries appear applies most severely to informational queries, which is the majority of publisher and affiliate traffic. Publishers built on answering questions face a mechanism that answers those questions without the visit, using their content as the input.

The strategic responses available fall into four groups, and none of them is comfortable.

Direct audience relationships. Newsletters, apps, podcasts, subscriptions, memberships and events convert a search-dependent audience into an owned one. The economics are harder — the audience is smaller and requires ongoing value delivery — but they are not subject to ranking changes. Publishers that built newsletter and subscription businesses before 2024 entered 2026 in a materially better position than those that did not.

Content types that resist synthesis. Original reporting, proprietary data, testing and reviews with genuine hands-on evidence, interviews, and opinion with a recognisable voice. These get cited rather than replaced, and citation carries some brand value even without a click. This is the same conclusion the earlier analysis of compounding assets reached, arrived at from a different direction, which increases confidence in it.

Licensing and direct arrangements with model providers. A growing number of publishers negotiated content licensing agreements with AI companies through 2024 to 2026. These provide revenue decoupled from pageviews. Access to such deals is concentrated among large publishers with distinctive archives, which does nothing for the long tail.

Commerce and transaction revenue. Affiliate arrangements, direct product sales, and marketplace models capture value per transaction rather than per impression, which survives lower traffic if the remaining traffic is more purchase-intent. Affiliate publishers focused on comparison and buying-decision content are somewhat better placed than those focused on informational content, because decision queries trigger summaries less often and because the reader needs to reach a merchant regardless.

For affiliates specifically, the medium-term risk is different and sharper. If assistants begin recommending products directly with integrated checkout — and the direction of travel in agentic commerce points there, with Google executives describing agentic experiences as changing shopping from discovery to decisions — the comparison-site layer of the value chain becomes structurally redundant. The affiliate model depends on the consumer needing an intermediary to compare options. An assistant that compares options is that intermediary. Whether the platforms will monetise that position through advertising, commission, or both is the open commercial question of the next two years.

The lesson for non-publishers is worth extracting, because it is easy to read this section as somebody else’s problem. Any business whose marketing depends on ranking for informational queries is, to a lesser degree, in the publisher’s position. The exposure is proportional to the share of commercial outcomes that begin with an informational search, and most companies have never calculated that share.

Regulated industries and the compliance tax

Financial services, pharmaceuticals, healthcare, gambling, alcohol, legal services, insurance and children’s products all operate under advertising rules that change the paid-versus-organic calculation in ways generic advice ignores.

Paid media in these categories carries a compliance overhead that raises its true cost well above the media price. Ad copy requires legal review. Claims require substantiation. Some claims are prohibited outright. Platforms impose their own restrictions, often broader than the law requires and applied by automated systems that reject compliant ads and approve non-compliant ones with roughly equal frequency. Certification requirements apply in several categories — financial promotions, pharmaceutical advertising, gambling operators — and add lead time and administrative cost. A campaign that takes two days to launch in retail takes six weeks in regulated finance, which removes most of paid media’s latency advantage.

Targeting restrictions compound this. Platforms restrict targeting by health condition, financial status, and other sensitive categories, which removes the precision that justifies paid social’s price in other sectors. Retargeting is constrained where it would reveal a sensitive interest. Lookalike modelling from customer lists raises data protection questions that legal teams in regulated sectors take seriously.

Organic content in these sectors faces the opposite pattern: higher production cost, lower velocity, but greater durability and better strategic fit. Content requiring medical, legal or financial review is expensive and slow to produce. It is also difficult for competitors to replicate, tends to rank durably because ranking systems in sensitive categories weight expertise and trustworthiness signals heavily, and is the natural source material for assistants answering questions in domains where accuracy matters. Google’s own quality guidance has long treated topics affecting health, financial stability, safety and legal rights as requiring higher demonstrated expertise, which means the compliance cost that makes content expensive is the same cost that makes it defensible.

The practical implication is that regulated industries have a stronger structural case for organic and owned investment than most sectors, and a weaker case for aggressive paid experimentation. The compliance overhead is a fixed cost per asset. Amortising it across a durable content asset that performs for three years is better arithmetic than amortising it across an ad creative with a six-week life.

Two further considerations apply. Consent and data protection constraints hit paid media harder than organic, because paid depends on tracking and organic depends on publication. In European markets with strong regulatory enforcement, the measurable portion of paid performance is smaller, which makes paid budgets harder to defend internally. And reputational risk asymmetry matters: a non-compliant ad in a regulated sector can trigger regulatory action, whereas a cautious content asset rarely does.

Gambling and alcohol face an additional dimension, which is the possibility of paid channels being closed by regulation entirely. Several European jurisdictions have restricted or banned gambling advertising in particular media and time slots, and further restriction is a live policy question in multiple markets. A business whose acquisition depends on a channel that may be legislated away has a strategic reason to build organic and owned distribution that has nothing to do with cost per acquisition. That is a risk-management argument rather than an efficiency argument, and it is the strongest form of the organic case in these sectors.

Budget size changes the correct answer

Advice about media allocation is usually written for companies with more money than the companies reading it. The correct split at €3,000 a month is not a scaled-down version of the correct split at €300,000 a month, because several mechanisms in marketing have minimum viable thresholds.

Below roughly €3,000 a month in total marketing investment, most conventional advice is inapplicable. Paid media at this level cannot buy enough frequency for brand effects and cannot generate enough conversions for the platform’s models to learn, which means automated bidding performs poorly and manual management consumes disproportionate time. Content production at this level supports perhaps one substantial piece a month. The rational concentration is on the narrowest possible high-intent capture — a small number of commercial search terms, a properly built local or category profile, review generation — plus whatever owned-audience mechanism fits the business. At this budget, the answer is neither paid nor organic in the strategic sense; it is capture plus retention, executed simply.

Between roughly €3,000 and €15,000 a month, both channels become viable but neither can be run at scale simultaneously. This is the range where the most damage is done by attempting a balanced plan, because a balanced plan at this level produces a paid programme too small to learn from and a content programme too small to rank. The higher-return approach is sequential concentration: pick the channel with the shorter payback given the business’s cash position, run it properly for two to three quarters, then use its returns to fund the second. For businesses with immediate cash needs that means paid first. For businesses with runway and a long sales cycle it frequently means content first.

Between roughly €15,000 and €75,000 a month, genuine portfolio management becomes possible and measurement becomes affordable. This is where geo experiments, dedicated conversion-rate work, a real content programme with distribution, and multi-platform paid activity can coexist. It is also where the 60/40 floor test starts to apply in practice, because the brand-building portion is large enough to produce a detectable effect.

Above roughly €75,000 a month, the constraint shifts from money to organisational capability. Marketing mix modelling becomes worth building. Broad-reach paid brand building becomes viable. Original research programmes, video production with recognisable presenters, and community investment can be properly funded. The binding constraint becomes the quality of decision-making and the honesty of measurement rather than the size of the budget.

Allocation starting points by budget level and business type

These are starting positions for a first quarter, not recommendations. Every row should be replaced within two quarters by whatever the company’s own experiments and modelling indicate, and the ranges assume a functioning product, a converting website and honest reporting. Companies missing any of those three should fix them before adjusting any percentage, because allocation improvements are worth a fraction of what conversion-rate and offer improvements are worth.

One further budget consideration is frequently missed. The correct split depends on the company’s cash position and cost of capital, not only on expected returns. Organic content is an investment with a two-year payback. A company with cheap capital and patient shareholders can fund it. A company that needs revenue in ninety days cannot, regardless of what the long-run arithmetic says. Recommending content-led growth to a business with four months of runway is bad advice even when the content arithmetic is superior, and recommending paid-led growth to a well-capitalised business in a category with durable search demand wastes an advantage.

Unit economics decide the split, not preference

The most reliable way to reach a defensible allocation is to work backwards from unit economics rather than forwards from channel philosophy. Four numbers determine the answer, and most companies can calculate all four in an afternoon.

Contribution margin per customer. Revenue minus cost of goods, fulfilment, payment processing, returns, and direct cost to serve, over the expected relationship rather than the first transaction. This sets the ceiling on what any acquisition can cost.

Payback period tolerance. How many months the business can wait to recover acquisition cost, determined by cash position, financing cost and growth expectations. A business that must recover acquisition cost within one month is restricted to capture channels. A business that can wait eighteen months has access to the entire portfolio.

Category demand elasticity. Whether marketing can expand total demand or only redistribute existing demand. Emergency services, replacement parts and prescription categories have near-fixed demand. Discretionary consumer goods, new software categories and lifestyle products have expandable demand. Fixed-demand categories should weight toward capture; expandable-demand categories have a real return on creation activity that fixed-demand categories do not.

Existing brand strength. Measured most practically by branded search volume relative to category search volume, plus unaided awareness where a company can afford to measure it. A brand already receiving substantial branded search has demand to capture; a brand receiving none has demand to create, and paid capture channels will deliver little because nobody is searching for it.

Combining these produces decision rules that are more useful than percentage recommendations.

When contribution margin is thin and payback tolerance is short, weight heavily toward capture and retention. Demand creation cannot be afforded, and the priority is extracting more value from existing demand through conversion-rate work, average order value, and repeat purchase.

When contribution margin is healthy and payback tolerance is long, weight toward creation. This is the position of well-funded B2B software, professional services with high deal values, and premium consumer brands. Capture channels will saturate quickly because the in-market population is small, and further growth requires expanding the population that considers the brand.

When branded search is negligible relative to category search, creation is the only route to growth, because there is nothing to capture. Companies in this position that pour budget into paid search on generic terms compete against established brands with better conversion rates and lose the auction on economics.

When branded search is large relative to category search, capture is cheap and creation should be evaluated against retention. A brand with strong demand often gets more from improving retention and average order value than from creating additional demand.

When demand is fixed and share is the game, competitive capture and conversion quality decide outcomes, and demand-creation spending mostly benefits competitors who capture the demand more cheaply.

The reason this framework outperforms channel-first thinking is that it produces different answers for different businesses, which is the correct behaviour for a decision that genuinely depends on circumstances. Any framework that produces the same recommendation for a plumber and a Series B software company is not a framework; it is a preference wearing one.

Consent rules reshaped both sides of the plan

Data protection regulation is usually discussed as a paid media problem. It is also an organic and owned media problem, and the second effect is larger than most companies have registered.

The European position is the strictest and therefore the most instructive. The General Data Protection Regulation requires a valid legal basis for processing personal data, and the ePrivacy Directive requires consent before storing or accessing information on a user’s device for non-essential purposes. In practice, this means advertising and analytics cookies require consent, and consent rates vary substantially by market, sector and consent-banner design. The commercial consequence is that a large share of European web sessions is invisible to both advertising platforms and analytics systems, which degrades paid targeting, paid measurement and organic performance reporting simultaneously.

The Digital Markets Act added obligations for designated gatekeepers, including restrictions on combining personal data across services without consent. Meta’s response — a consent-or-pay model offering either personalised advertising or a paid subscription — drew a €200 million fine under the DMA, and through 2025 and 2026 the company moved toward offering EU users a less personalised advertising option while publicly defending personalised advertising as a model. The direction of travel in Europe is toward less personal data available for paid targeting, which mechanically reduces the precision advantage that justified paid social’s pricing.

Google’s reversal on third-party cookies removed one expected disruption and created a different kind of uncertainty. After years of announced deprecation timelines, Chrome retained third-party cookies and Google wound down the Privacy Sandbox APIs. The practical effect is not a return to 2018 conditions, because signal loss had come primarily from other browsers’ default blocking, from consent frameworks, and from platform policy rather than from Chrome. Companies that used the cookie deprecation timeline as the reason to build first-party data capability were right for the wrong reason, and the capability remains correct.

The consequences for allocation are specific.

Paid media becomes more expensive per unit of measured outcome in high-consent-friction markets, because a portion of conversions cannot be observed and must be modelled, and because targeting precision falls. This is not a reason to abandon paid media; it is a reason to expect European paid performance metrics to understate actual performance, and to rely more heavily on geo experiments and modelling, both of which are unaffected by consent because they use aggregate data.

Organic and owned channels become relatively more attractive on measurement grounds specifically, because a newsletter subscriber who consented to marketing email is a fully addressable, fully measurable relationship that no browser change affects. First-party data collected with genuine consent is the only marketing asset whose measurability is improving rather than degrading.

Content strategy acquires a compliance dimension that is easy to overlook. Personalisation of on-site content, behavioural email triggering, and lead scoring all process personal data and require legal bases and transparency. Companies that build sophisticated content personalisation without corresponding governance accumulate risk that surfaces during due diligence or a data subject access request.

There is a second-order effect worth naming. Consent friction pushes companies toward channels where the platform holds the consent relationship rather than the advertiser. When tracking a user across a company’s own site requires consent that many users decline, but a platform can serve ads inside its own logged-in environment on the basis of consent it obtained, the platform’s owned inventory becomes relatively more attractive. Regulation intended to reduce tracking has, in this specific respect, strengthened the position of the largest platforms against both independent publishers and independent advertisers. That is an interpretation of the competitive effect rather than a stated policy objective, but the mechanism is straightforward.

Platform dependency as a balance-sheet risk

Marketing plans are evaluated on returns and rarely on concentration risk, which is odd given that concentration risk is standard analysis in every other area of business.

Consider how a procurement function would assess a supplier relationship with these properties: the supplier sets prices unilaterally, has raised them roughly 10% to 15% annually for a decade, controls the data used to evaluate its own performance, competes with the customer for end-customer relationships, can change delivery terms without notice, can terminate the relationship without appeal, and has two or three viable alternatives at most. No competent procurement function would accept that arrangement for a critical input, and most companies accept it for the majority of their customer acquisition.

The specific dependency risks are enumerable rather than vague.

Price risk is the one already visible. Meta’s 12% year-over-year increase in average price per ad in the first quarter of 2026 is a supplier exercising pricing power. Absent margin expansion elsewhere, the customer absorbs it.

Policy risk is the sudden loss of a tactic. Categories get restricted, targeting options are removed, ad formats are deprecated, and account suspensions happen with limited recourse. Businesses in supplement, cryptocurrency, adult, weapons-adjacent, political and certain health categories have experienced complete channel closure with days of notice.

Algorithm risk applies to organic distribution and is the mirror image. Search ranking updates have removed a majority of a site’s traffic overnight. Social platforms have reduced reach for link posts, for external content, and for entire content formats. A company whose revenue depends on ranking is exposed to a supplier who provides no contract, no notice and no compensation.

Disintermediation risk is the most strategically serious and the least discussed. Platforms increasingly transact directly. Marketplace models, in-app checkout, agentic shopping and platform-native commerce all move the customer relationship to the platform. A business that acquires customers through a platform that then begins fulfilling the transaction has been reduced to a supplier.

Measurement dependency risk compounds all of the above. When a company cannot evaluate its own marketing independently, it cannot detect when the supplier’s performance degrades, and it cannot negotiate.

The mitigations are unglamorous and cumulatively substantial. Maintain at least three material acquisition channels, defined as channels that could each carry 25% of new customers if the others were unavailable. Build an owned audience with a size target and an accountable owner. Hold independent measurement capability, at minimum a reconciliation of platform-reported conversions against actual orders and a periodic geo experiment. Keep first-party data in systems the company controls rather than only inside platform environments. Retain the operational capacity to sell directly even where most volume flows through intermediaries.

None of these mitigations improves next quarter’s return on ad spend, which is precisely why they are underfunded. They are insurance, and their value shows up only in the scenario they insure against. The organic and owned side of the marketing plan is, viewed this way, partly a hedge rather than a growth channel — and hedges are properly evaluated on the size of the exposure they cover rather than on their standalone return.

The labour cost nobody puts in the media plan

Paid media budgets are reported as media spend. Organic content budgets are frequently reported as agency fees or tool subscriptions, with the largest input — people — sitting in a different cost line. That accounting difference distorts every comparison a company makes between the two.

Gartner’s 2026 CMO Spend Survey found labour consuming 24.5% of marketing budget, up from 21.9% in 2025. That is the reported marketing labour figure, and it understates the content-relevant portion because much content work is performed by product, sales, engineering and executive time that never appears in a marketing cost centre. A founder spending six hours a week on original commentary and video is contributing a substantial content investment that appears nowhere in the marketing budget.

The correct comparison requires fully loaded costs on both sides. A content programme producing eight substantial pieces a quarter with two full-time equivalents at a fully loaded cost of €70,000 each is a €140,000 annual programme regardless of how it is booked, plus tools, plus freelance production, plus the distribution effort. Compared honestly against €140,000 of paid media, the content programme’s apparent cost advantage narrows considerably, and its payback timeline is much longer. Many content programmes that look cheap look that way only because their main input is not being counted.

The same discipline applies to paid media, where it usually cuts the other way. Paid media requires skilled management, creative production, landing page work, analytics and reporting. An account spending €40,000 a month with an agency fee of €5,000 and internal creative and analytics support has a true cost well above the media number, and the incremental cost of managing a larger budget rises slowly, which is a genuine scale advantage for paid over content. Content cost scales roughly linearly with output; paid management cost scales sub-linearly with spend. That asymmetry is one of the strongest structural arguments for paid at large budget levels and against it at small ones.

Two labour dynamics changed materially by 2026. The first is that generative tools reduced the cost of adequate content production sharply while leaving the cost of distinctive content roughly unchanged. Drafting, editing, translation, repurposing, transcript-to-article conversion and basic image production became substantially cheaper. Original research design, expert interviewing, opinion with reputational stakes, on-camera presence, and community management did not. The labour saving landed precisely on the content type whose distribution value fell, which is a poor coincidence for content teams hoping tools would solve their economics.

The second is the capability gap the Gartner research documents. Seventy percent of CMOs reported their internal processes were not mature enough for AI implementation, with only 30% describing themselves as ready to scale, and 38% citing lack of internal AI expertise as the top barrier. Ewan McIntyre, VP Analyst and Chief of Research in Gartner’s marketing practice, framed the underlying point directly: “AI changes the kind of capability needed, but doesn’t eliminate capability itself.”

That has an allocation consequence. A company investing 15.3% of marketing budget in AI while lacking the process maturity to use it is funding capability it cannot yet deploy, and that money comes from somewhere. In practice it has come partly from the middle of the funnel — the Gartner finding that loyalty and retention spending fell 29% since 2024 while awareness and conversion consumed 62.6% of media is consistent with budget being pulled from the least measurable activities to fund the most fashionable one.

The practical guidance is simple and rarely followed: build a single fully loaded cost view of all marketing activity, including internal time at fully loaded rates, and rerun the paid-versus-organic comparison on that basis before making any allocation decision. A large share of companies that do this for the first time discover their content programme is more expensive than they believed and their paid programme is less profitable, which is uncomfortable and useful.

AI production economics and the content flood

Generative tools changed the supply curve for content, and the consequences follow from basic economics rather than from anything specific to the technology.

When the marginal cost of producing adequate content falls toward zero, the quantity supplied rises sharply. The quantity demanded by distribution systems does not rise, because attention is fixed and ranking positions are fixed. The result is a supply glut in which the clearing price of adequate content approaches zero and the premium on genuinely scarce content rises. That is the whole story, and most content strategy debate in 2026 is a rediscovery of it.

Several observable effects follow. Search ranking systems face vastly more candidate documents of similar surface quality, which increases the weight placed on signals that are hard to fabricate: independent citation, brand recognition, demonstrated expertise, first-hand evidence, and user behaviour. Publishing volume stopped being a competitive advantage and became a cost. Content that could have been generated by a model competes against content that was generated by a model, at a much lower cost base, from thousands of competitors simultaneously.

The Content Marketing Institute research captured how thoroughly this became standard practice: 95% of B2B marketers reported using AI-powered applications, 89% for content creation, with 87% reporting productivity improvement and 58% reporting quality gains. Implementation maturity clustered in the middle — 20% exploratory, 48% developing, 24% established, 5% advanced, 3% leading. When 89% of a profession uses the same tools for the same task, the tools cannot be a source of advantage. They are table stakes, and the advantage moves to whatever they cannot do.

What they cannot do, at least presently, defines the content investment worth making: run an experiment, survey a population, analyse a proprietary dataset, hold an opinion with professional consequences attached, appear on camera as a person the audience recognises, interview a named expert, moderate a community, build a tool, or state what happened when a specific approach was tried on a specific account. Everything in that list requires either proprietary access, personal reputation, or real-world action, and none of it can be generated.

There is a second-order effect on paid media that deserves attention. Cheaper creative production increased the volume of advertising creative in circulation, which raised the frequency at which audiences encounter advertising and, plausibly, accelerated creative fatigue. Platforms simultaneously deployed generative creative tools that produce variations automatically. The result is more advertising, more similar advertising, and a higher bar for creative distinctiveness to earn attention. Cheap creative production does not reduce the cost of attention; it increases the competition for it. This mirrors precisely what happened on the organic side.

A third effect concerns the training and grounding data of the models themselves. As generated content fills the web, models increasingly encounter their own outputs as source material. The technical literature has flagged degradation risks from training on synthetic data, and grounding systems that retrieve from a corpus increasingly composed of generated summaries face a related problem. The practical implication for marketers is that being a primary source — the origin of a fact, figure, or dataset rather than a restatement of one — is likely to be worth more rather than less. That is a forecast rather than an established finding, and it is consistent with the direction of the incentives on all sides.

The allocation conclusion from all of this is not that content investment should fall. It is that content investment should become more concentrated, more expensive per unit, more original, and evaluated over longer horizons — while the routine content production that occupied most content budgets moves toward being a low-cost utility function rather than a strategy.

Building the hybrid system that actually works

The conclusion that companies should do both is true and nearly useless on its own, because doing both badly is the most common outcome. What separates a functioning hybrid from parallel mediocrity is a small number of specific mechanisms, and they are operational rather than conceptual.

Shared audience definition. Paid targeting and content topic selection should derive from the same understanding of who the customer is and what they are trying to do. In most companies they do not: paid targeting comes from platform audience tools and past performance, content topics come from keyword volume data. The two teams work from different maps of the same territory. Building one document that defines segments, their situations, their questions at each stage, and their objections — then requiring both teams to reference it — resolves a surprising number of downstream conflicts.

Paid as a content testing instrument. Before committing a quarter of content production to a theme, test the theme’s resonance with a few hundred euros of paid distribution. Message tests, hook tests, and offer tests return usable signal in days. Content teams that use paid budget as a research instrument produce better content and can justify their choices with evidence rather than keyword volume. This is the single highest-return integration point between the two disciplines and it is rarely implemented, because the budgets sit in different places.

Content as paid creative supply. The best-performing organic content — the posts that earned reach, the videos that held attention, the research that got cited — is proven creative. Promoting proven organic content is materially lower-risk than producing new advertising creative on a hypothesis. Companies that run a systematic pipeline from organic performance data into paid creative selection get better paid results at lower creative production cost.

Paid support for organic assets. A piece of original research does not distribute itself. Spending €3,000 promoting a €15,000 research study to a precisely defined professional audience frequently produces more citation, more links, more coverage and more inbound interest than the study would achieve organically over a year. Treating paid distribution as part of the content production budget rather than as a competing line item is the accounting change that makes this happen.

Organic capture of paid-created demand. Paid brand campaigns create branded search demand. If the branded search results page is weak — no strong organic listing, poor comparison content, competitor ads above the brand — the demand leaks. Ensuring the organic and owned surfaces are ready to receive paid-created demand costs almost nothing and recovers a share of paid spend that would otherwise be wasted.

One measurement framework covering both. Separate reporting for paid and organic guarantees the political conflict continues, because each team reports the numbers that flatter it. A single framework — total blended acquisition cost, contribution margin, incremental revenue from tested activity, branded search volume, owned audience growth, and pipeline or revenue by cohort — forces both teams to argue against the same evidence.

A single accountable owner for the split. Where paid and organic report to different executives, the split is decided by negotiation rather than analysis. Where one person owns both and is accountable for blended acquisition cost and total revenue growth, the incentive to allocate honestly exists. This is an organisational design decision that determines whether every mechanism above is achievable.

A standing experiment calendar. One incrementality test per quarter, rotating through the tactics most likely to be overstated: brand-term bidding, retargeting, prospecting, content-driven activity, email frequency. Four tests a year across three years produces a genuine map of what works in a specific business, which is worth more than any benchmark.

The system described here is not expensive relative to the budgets it governs. Its main cost is organisational discomfort: it requires the paid team to expose its incrementality, the content team to expose its production economics, and the executive to fund activity with a two-year payback while under quarterly pressure. Companies that get the allocation right are usually not the ones with better analysis. They are the ones with the organisational structure that permits honest analysis to change decisions.

A 90-day plan for reallocating budget

The following sequence assumes a company already spending on both paid media and content, with reporting that does not clearly answer whether the split is right. It is ordered by information value per unit of effort rather than by conceptual tidiness.

Days 1 to 10: build the honest cost and outcome baseline. Assemble total marketing investment for the past twelve months with fully loaded costs: media spend, agency fees, tools, freelancers, and internal salaries apportioned by time. Assemble new customers, revenue, contribution margin, and retention by cohort for the same period. Calculate blended acquisition cost by month and paid-only acquisition cost by month, and plot both. Sum platform-reported conversions across all advertising platforms and divide by actual new customers to establish the attribution inflation ratio. A ratio above 1.5 means channel-level reporting cannot be used for allocation decisions, and most companies find a ratio between 1.4 and 2.5.

Days 11 to 20: audit the content library by query type and commercial function. Classify every indexed page by intent — informational, comparative, transactional, navigational, retention — and by traffic and conversion contribution over twelve months. Identify the share of production investment sitting in informational content on question-shaped queries, which is the category most exposed to generative summarisation. Identify pages with traffic and no conversion path, and pages with conversion intent and no traffic. This audit routinely reveals that 60% to 80% of pages contribute almost nothing, and that the conversion-relevant pages are under-invested.

Days 21 to 30: run the assistant visibility check. Write 30 to 50 prompts representing real buying questions in the category, including comparison prompts, alternatives-to-competitor prompts, and problem-framing prompts. Run them across the major assistants, record whether the brand appears, how it is described, what it is compared against, and which sources are cited. Repeat monthly. This costs one person one day per month and is currently the only reliable read on answer-layer position.

Days 31 to 45: design and launch the first incrementality test. Choose the tactic with the highest reported return and the highest suspected overstatement, which is usually brand-term bidding or site retargeting. For brand terms, pause bidding in a matched set of regions or on a matched share of traffic for at least four weeks and measure total conversions, not attributed conversions. Ensure the test period avoids promotional distortion and that the control and test regions are matched on historical trend rather than on size alone.

Days 31 to 60, in parallel: fix conversion and offer before touching allocation. Review the top five landing destinations for paid traffic and the top five for organic traffic. Test one substantial change on each: clearer offer, reduced form friction, added proof, better pricing transparency, faster load. Conversion-rate improvements apply to every channel simultaneously and therefore outrank allocation changes in return per hour of effort.

Days 46 to 60: build the owned-audience capture mechanism and set a target. Identify every point where a visitor demonstrates interest without giving a means of contact, and add one. Set a numeric monthly target for list growth with a named owner. Report it alongside acquisition metrics.

Days 61 to 75: reconstruct the content plan around what cannot be synthesised. Cancel planned content that a language model could produce from public information. Replace it with a smaller number of items that require proprietary data, first-hand testing, named expertise, customer evidence, or tooling. Allocate 40% to 50% of each item’s total budget to distribution before production begins.

Days 76 to 90: read the test, reallocate, and set the standing cadence. The incrementality test should now have a readable result. If the paused tactic showed low incrementality, redirect that budget — first to conversion work, then to the content and owned programmes, then to prospecting. Schedule the next quarter’s test. Agree with the executive team the evaluation horizon for the content programme, in writing, and the intermediate metrics it will be judged on before revenue appears.

Two cautions. Nothing in this sequence requires a large budget, and nothing in it produces a visible result inside 90 days except the conversion work and the test result. Companies expecting the reallocation itself to change revenue within a quarter will abandon it at the point of maximum cost and minimum return, which is the same failure described earlier in a different form.

The measurement stack worth paying for

Measurement spending is where marketing budgets waste money most quietly, because tools are easier to buy than analysis and dashboards feel like progress. A defensible stack for a mid-sized company has five layers, and the order matters more than the vendors.

Layer one: reliable outcome data. Actual orders, actual revenue, actual contribution margin, actual retention, by cohort, in a system the company controls. This sounds trivial and is frequently the broken layer. Companies with revenue data spread across a commerce platform, a payment processor, a CRM and a spreadsheet cannot measure anything reliably regardless of what they layer on top. No measurement investment above this layer returns anything until this layer is correct.

Layer two: server-side event collection with consent management. Collecting events on the company’s own infrastructure, respecting consent, and forwarding to platforms via conversion APIs. This restores a portion of the signal lost to browser restrictions and improves platform model quality, which improves automated bidding. It is technical work with unglamorous returns and it is worth doing before buying any analytics product.

Layer three: platform attribution, used for its actual purpose. Platform-reported data is appropriate for within-platform decisions: which creative, which audience, which keyword, which bid. It is not appropriate for deciding how much total budget a platform receives. Using it for the first and refusing to use it for the second is the whole discipline.

Layer four: experiments. A standing cadence of geo tests and holdouts, one per quarter minimum. This is where the actual answers come from. The cost is the withheld spend and the analytical time, and the return is the elimination of non-incremental spending, which in most accounts is a double-digit percentage of the paid budget.

Layer five: modelling. Marketing mix modelling once there are two or more years of weekly data with genuine spend variation, calibrated against the experiment results from layer four. Open-source options including Google’s Meridian and Meta’s Robyn removed the licence cost barrier; Google’s Meridian Studio announcement in May 2026 signalled the move toward continuously refreshed enterprise modelling, paired with Meridian GeoX for experimental calibration.

Three items commonly bought and rarely worth the money at mid-market scale: multi-touch attribution platforms that promise to resolve cross-channel credit at user level, which the underlying signal degradation has made structurally unreliable; dashboarding tools purchased before the data underneath them is trustworthy, which produce confident-looking wrong answers faster; and brand-tracking studies at sample sizes too small to detect the changes a mid-sized budget produces.

Two items commonly skipped and worth the money: someone whose job is analysis rather than reporting, and a small monthly budget for primary research asking customers how they found the company and what influenced them. Self-reported attribution is methodologically weak and directionally useful, and it is the only instrument that captures assistant-mediated and word-of-mouth influence at all. A single question at checkout or on the demo form, answered by a few hundred customers a quarter, frequently reveals channels that appear nowhere in analytics.

The final component is a reporting convention rather than a tool: report every channel with an explicit statement of the evidence tier behind its numbers. A line that reads “paid social: €18,000 spend, 220 conversions, platform-attributed, untested” and a line that reads “paid search brand terms: €6,000 spend, 310 conversions, platform-attributed, geo test showed 22% incrementality” communicate radically different confidence with no additional analysis. Companies that adopt this convention find their allocation arguments shorten considerably.

Scenarios for the next 24 months

Forecasting in this area has a poor record, so the honest form is a small number of scenarios with their observable early indicators rather than a single prediction. What follows is analysis and interpretation, not measurement.

Scenario one: gradual absorption. Assistant and generative surfaces continue growing, click-through from classical search continues declining slowly, and platforms monetise the new surfaces progressively. Paid media prices continue rising in the 8% to 15% annual range as the new inventory is thin and expensive. Organic traffic to most commercial sites declines modestly each year while assistant-referred traffic grows from a small base at high conversion rates. Companies adapt by shifting content toward original material and owned audiences. This is the extension of the observable 2025 to 2026 trend and, in this analysis, the most likely path. Early indicators: continued single-digit decline in classical organic sessions, continued double-digit growth in assistant referrals, steady paid price inflation, no discontinuity.

Scenario two: monetised answer layer at scale. Ads inside AI answers become a major inventory class faster than expected. Google’s stated capability to monetise longer and more complex queries, plus OpenAI’s advertising build-out, converge into a large new paid surface. This would reduce the answer layer’s status as an unbuyable organic opportunity and reprice it as an auction. For advertisers this means new inventory at high prices and a further squeeze on organic visibility, because paid answers occupy the shortlist positions. Early indicators: rapid expansion of ad load inside assistant answers, dedicated buying interfaces rather than eligibility flowing from existing campaigns, disclosed revenue from these surfaces in quarterly reporting.

Scenario three: agentic commerce disintermediation. Assistants move from recommending to transacting. Agentic checkout, integrated payments, and platform-brokered purchasing become common in defined categories. Google executives have described agentic experiences as changing shopping from discovery to decisions, and retailer integrations with both Google and OpenAI began appearing through 2026. In this scenario the comparison and affiliate layer contracts sharply, brand recognition matters more because it determines which options an agent considers, and price and availability data feeds become critical marketing infrastructure. Early indicators: transaction volume disclosed through assistant checkout, retailer feed integration requirements, changes in affiliate network revenue.

Scenario four: regulatory disruption. Antitrust remedies, data protection enforcement, or AI-specific regulation materially change platform behaviour. The US search antitrust remedies process, with the district court’s remedies ruling and subsequent appeal activity through 2025 and 2026, and continued DMA enforcement in Europe, both create paths to changed default arrangements, changed data-combination rules, or changed self-preferencing behaviour. Effects on advertisers are hard to predict and could reduce or increase concentration. Early indicators: enforced changes to default search arrangements, mandated data access or portability, further consent-model rulings.

Scenario five: reversion. Users find generative answers insufficiently trustworthy for commercial decisions and return to evaluating sources directly, or platforms pull back on summarisation for commercial queries because it damages advertising revenue. This scenario receives little attention and is not implausible: the entity with most to lose from zero-click behaviour on commercial queries is the company selling the clicks. Early indicators: reduced summary trigger rates on commercial and transactional queries, recovery in click-through rates, publisher traffic stabilisation.

The practical value of holding all five is that the actions which look sensible across most of them are the actions worth taking now. Original content that cannot be synthesised is useful in scenarios one, two, three and five. Owned audience is useful in all five. Independent measurement is useful in all five. Reduced dependency on any single channel is useful in all five. A strategy that only works in one scenario is a bet; a strategy that works in four is a plan.

Open questions the evidence cannot settle yet

Intellectual honesty requires marking where the analysis runs out. Six questions are commercially important and currently unanswerable from available evidence.

How much commercial value does uncited or unclicked assistant exposure carry? When a model describes a brand favourably without a citation or a visit, something has happened. It resembles brand advertising delivered by an intermediary. Nobody has a measurement method for it, and its magnitude could be trivial or larger than the traffic effects everyone is discussing. Until self-reported attribution studies or brand-tracking work at adequate scale addresses it, this is a genuine unknown at the centre of the whole question.

What is the durable conversion-rate advantage of assistant-referred traffic? The reported multiples of roughly four to five times organic conversion rates come from small samples with severe self-selection: current assistant users skew toward earlier-adopting, higher-intent, more technically confident populations. Whether the advantage persists as usage broadens to the general population is unknown, and there is a reasonable prior that it will compress.

Where does paid media price inflation stop? Prices have risen consistently for a decade. Either advertiser returns are still adequate at current prices, in which case inflation continues until they are not, or a substantial cohort of advertisers is mismeasuring and will withdraw when they discover it. The absence of a visible ceiling does not mean there is not one, and the shape of the demand curve above current prices has never been observed.

Does original, expensive content actually earn a durable premium in assistant citation? The argument is logical and the incentives point that way, but the empirical base is thin. Citation patterns could equally settle on a small number of high-authority general sources — encyclopaedic references, large publishers, dominant review platforms — leaving specialist original work unrewarded. Pew’s finding that Wikipedia, YouTube and Reddit supplied 15% of summary sources is at least consistent with concentration.

How will regulation change the competitive position of independent publishers and advertisers? Enforcement outcomes in the US search antitrust case and under the DMA remain in progress, and their effects on advertising markets specifically have not been modelled convincingly by anyone.

Is the 60/40 brand-to-activation heuristic still approximately right? The measurement asymmetry it corrects for has grown worse while the cost of brand building has risen. Those forces push in opposite directions and no updated analysis on a comparable case base has resolved the net effect.

Anyone claiming confident answers to these six is extrapolating. The correct posture is to build a marketing system that performs acceptably under several answers rather than optimally under one, and to keep measuring, because these questions will be settled by data that does not exist yet.

The verdict, expressed as a decision rule

The question as posed has no general answer, and the specific answer is derivable. Here is the derivation, stated as compactly as the evidence permits.

Paid media is better when timing has value, when demand already exists in measurable volume, when the business needs revenue faster than content can compound, when unit economics tolerate the auction price, and when the activity being funded is genuinely incremental rather than harvesting demand created elsewhere. Those conditions describe launches, seasonal peaks, market entry, testing, high-margin businesses with short payback requirements, and categories where in-market intent is searchable and abundant.

Organic content is better when demand must be created rather than captured, when the business can wait through a twelve to twenty-four month payback, when the category rewards expertise that cannot be generated, when the content produces something a model cannot synthesise, and when the alternative is renting the same attention permanently at rising auction prices. Those conditions describe B2B software, professional services, considered consumer purchases, regulated sectors, and any business with the capital to build assets rather than lease distribution.

Both are wrong when the underlying problem is conversion rate, offer quality, product-market fit, retention, or pricing. A large share of allocation arguments are displacement activity for a problem that no media split solves, and the diagnostic is simple: if the website converts poorly, if customers do not return, or if the offer is undifferentiated, allocation is the fourth-most-important decision on the list.

Three rules survive scrutiny across every category examined here.

First: the split is an empirical question with a local answer, and it should be settled by experiment rather than argument. One incrementality test per quarter, rotating through the tactics most likely to be overstated, produces more decision value than any amount of benchmarking. Companies that run them consistently find a portion of paid spend is not incremental and a portion of content investment is more productive than reported.

Second: any plan allocating less than roughly 30% of total marketing investment to demand creation and owned-audience building is probably under-invested there, and the measurement system will never tell you. This is the defensible remainder of the 60/40 evidence after its limits are acknowledged, and it holds because the measurement bias it corrects for got worse rather than better.

Third: the strategic objective is not a ratio but reduced dependency. A business with three viable acquisition channels, an owned audience of real size, independent measurement, and content assets that cannot be synthesised has options. A business with excellent return on ad spend on two platforms it does not control has a good quarter and a fragile position. Meta raising unit prices 12% in a year, Google Network revenue falling 4% while Google’s owned Search grew 19%, and half of search clicks disappearing behind a generated paragraph are all the same event viewed from different angles: the owners of distribution capturing more of the value, and the businesses that depend on them absorbing the cost.

The companies handling 2026 well are not the ones that picked a side. They are the ones that stopped treating the question as a matter of belief, priced both options honestly including labour, tested what was actually incremental, and put the savings into assets nobody can reprice on them. That is a slower answer than the question invites, and it is the one the evidence supports.

Creator partnerships sit in neither column

Creator and influencer arrangements are the clearest demonstration that the paid-organic binary is a reporting convention rather than a description of reality, and they have become large enough that mis-categorising them distorts whole plans.

A paid partnership with a creator involves money changing hands for distribution, which places it in paid media. The distribution itself is delivered by an algorithm responding to audience engagement, which places it in organic. The content is produced by a third party whose reputation carries the message, which places it closer to earned media. The output frequently becomes advertising creative used in paid campaigns, which places it in production. Four different budget lines have a legitimate claim on the same activity, and in most companies whichever team negotiated the deal books it, which makes the resulting reporting incomparable across periods.

The commercial case for the format rests on three properties that neither pure paid nor pure organic offers. Creator content carries trust that brand-produced content does not, because the audience relationship predates the commercial arrangement. It reaches audiences that are hard to buy directly, particularly younger consumers and specialist professional communities. And it produces creative that performs when subsequently promoted, which is why the highest-return use of the format is frequently not the organic post but the paid amplification of it with usage rights secured in the contract.

The failure modes are equally specific. Follower count is close to uncorrelated with commercial outcome, because reach is allocated by engagement prediction rather than audience size, and because relevance dominates scale in conversion. Reach delivered is unpredictable, which makes the format unsuitable for guaranteed coverage in a launch window unless paid amplification is contracted alongside it. Measurement is poor, since the traffic frequently arrives as direct or untracked and the influence often precedes the purchase by weeks. And rate inflation has been severe in the categories where the format works best, to the point where a mid-tier creator partnership in some verticals costs more than the equivalent reach bought directly, with none of the delivery certainty.

Two practices separate companies that get value from the format from those that do not. The first is contracting usage rights as standard, so every partnership produces reusable advertising assets rather than a single post. The second is treating creator relationships as long-run rather than transactional, because repeated association builds the credibility the format is bought for, and a single sponsored post from an unfamiliar account is close to indistinguishable from an ad with worse targeting.

The regulatory dimension deserves a line, because enforcement has tightened. Disclosure requirements for paid partnerships apply across major jurisdictions, platforms provide disclosure tooling, and consumer protection authorities in Europe and the US have pursued cases against both brands and creators for inadequate labelling. Undisclosed partnerships also carry a commercial risk that exceeds the regulatory one: the format’s entire value is trust, and concealment is the fastest way to destroy it.

Where this lands for allocation is straightforward. Creator partnerships should be budgeted as paid media with content production characteristics, measured with the same experimental discipline applied to any other paid channel, and contracted so the output feeds the paid creative pipeline. Filing them under organic because no impressions were purchased flatters the organic column; filing them under paid without capturing the reusable assets wastes most of what was bought.

Objections from both camps, answered directly

Any argument for a mixed portfolio attracts predictable objections from committed practitioners on both sides. The strongest versions deserve direct answers rather than dismissal.

“Organic is dead, the data proves it.” The click data supports a serious contraction in one specific organic channel — classical search on informational queries — and does not support the general claim. Search demand rose while click rates fell. Comparative, transactional and decision-support queries hold up better than informational ones. Owned email and community distribution are unaffected by any of it. Assistant referrals grow from a small base and convert well. The correct statement is that the traffic yield of one organic tactic collapsed, not that earned distribution stopped working.

“Paid media is a tax you pay forever, so it can never be the right answer.” Rent is also a payment that never builds equity, and renting is nonetheless correct for a business that needs premises next month and cannot buy a building. The permanence objection is a real consideration for a mature business with capital and time. It is irrelevant to a business whose alternative is having no distribution at all for eighteen months.

“Our platform-reported return on ad spend is 5:1, so paid clearly works.” Platform-reported return measures the platform’s account of its own contribution. The relevant test is what happens to total revenue when the spend is withdrawn. Companies that run that test frequently find a portion of the reported return was demand that would have converted anyway. Reported 5:1 is a hypothesis, not a finding, until a holdout has tested it.

“We tried content for six months and it produced nothing.” In a competitive category in 2026, six months is roughly a third of the way to a readable result, and the outcome of stopping at that point is guaranteed regardless of whether the programme was sound. The relevant questions are whether the content was distinguishable from what a model could generate, whether anything was spent distributing it, and whether the intermediate indicators — branded search, direct traffic, list growth, sales usage — moved. A programme with no distribution budget and no intermediate metrics did not test whether content works; it tested whether publishing alone works, and that question was already settled.

“AI makes content production so cheap that volume is back.” Cheap production increased supply without increasing the distribution available to absorb it, which lowered the return on adequate content rather than raising it. The Content Marketing Institute finding that 89% of B2B marketers use AI for content creation establishes that this capability is universal, and universal capabilities do not produce advantage.

“Incrementality testing is impossible at our size.” Below a few dozen conversions a month, formal testing is genuinely underpowered, and that is a real constraint rather than an excuse. The available substitutes are longer channel pauses of six to eight weeks with careful seasonality control, and tracking branded search volume and direct traffic as leading indicators of demand-creation health. Small businesses cannot measure precisely and can still avoid the two largest errors: paying for demand they already had, and cancelling demand creation before it could have worked.

“Our competitors are all increasing paid spend, so we must match them.” In an auction, matching competitors’ spending increases raises the clearing price for everyone and transfers margin to the platform. Where a category’s advertisers all escalate simultaneously, the differentiated position is frequently the one that competes on conversion rate, offer, retention and owned audience instead — because those levers are not priced by an auction and cannot be bid away.

Common questions about the paid and organic split

Author:
Jan Bielik
CEO & Founder of Webiano Digital & Marketing Agency

This article is an original analysis supported by the sources cited below

Google users are less likely to click on links when an AI summary appears in the results Pew Research Center analysis of 68,879 Google searches by 900 US adults during March 2025, providing the click-rate comparison between searches with and without AI summaries.

Gartner 2026 CMO Spend Survey finds CMOs allocate 15.3% of marketing budgets to AI Survey of 401 CMOs and marketing leaders across North America, the UK and Europe conducted January to March 2026, covering AI allocation, readiness and capability barriers.

Gartner marketing survey finds awareness and conversion account for 62.6% of total media spend Gartner data on media spend by objective, the 29% decline in loyalty and retention investment since 2024, and labour rising to 24.5% of marketing budget.

Gartner 2025 CMO Spend Survey reveals marketing budgets have flatlined at 7.7% of overall company revenue The prior year’s benchmark establishing that marketing budgets stopped growing as a share of revenue, which is the constraint behind internal budget conflict.

Meta reports first quarter 2026 results Primary financial disclosure containing the 19% increase in ad impressions, 12% increase in average price per ad, $56.31 billion revenue and 3.56 billion daily active people.

Alphabet announces first quarter 2026 results Alphabet’s quarterly filing with segment revenue for Google Search and other, YouTube advertising, and Google Network.

Alphabet earnings call, Q1 2026: Sundar Pichai’s remarks Google’s own account of AI Overviews and AI Mode driving increased query volume, used here for the claim that search demand rose while click rates fell.

Alphabet Q1 2026: Google Network ad revenue falls 4% as AI reshapes the web Trade analysis of the segment figures and executive commentary, including Philipp Schindler’s remarks on Gemini expanding monetisation of longer, more complex queries.

Google Marketing Live 2026: growth in the age of AI Google’s May 2026 announcement of Meridian GeoX geographic incrementality testing and Meridian Studio, plus measurement positioning from Gaurav Bhaya.

AI Mode is Google’s next ads engine — and it already knows how to monetize it March 2026 analysis of ad formats inside AI Mode, campaign eligibility routes, and the argument for gradual rather than aggressive monetisation.

WARC global ad forecasts upgraded but growth concentrated within Big Tech December 2025 forecast placing the global ad market at $1.19 trillion with 9.1% growth expected in 2026 and incremental spend concentrating in Alphabet, Amazon and Meta.

B2B content and marketing trends research Content Marketing Institute and MarketingProfs survey of 1,015 B2B marketers conducted June to August 2025, covering how marketers rate results, AI adoption, challenges and 2026 investment priorities.

The key works of Les Binet and Peter Field The IPA’s collection of the effectiveness research behind the 60/40 brand-building and activation split referenced throughout this analysis.

Les Binet on brand building: “60/40 is not an iron rule” Interview in which Binet sets out the conditions under which the ratio varies, used here to qualify how the heuristic should be applied.

How do you measure “How Brands Grow”? Ehrenberg-Bass Institute material on penetration, mental availability and reach, supporting the argument that growth depends on reaching light and non-buyers.

US ad spending 2026 Forecast context on channel growth, Meta’s projected US net digital ad revenue position, and the shift of spend toward social and retail media.

FAQ on incrementality: how to prove your ads actually work in 2026 Practitioner overview of holdout and geographic experiment design, supporting the evidence hierarchy set out in this article.

Meta’s 2026 DMA report reveals WhatsApp ads, a €200m fine, and a defiant stance on personalized advertising Documentation of the Digital Markets Act fine, Meta’s consent-or-pay position, and the move toward a less personalised advertising option for EU users.

Google Chrome is keeping third-party cookies after all: what does it mean? Consent-management analysis of Google’s reversal and why the underlying measurement signal loss came largely from other sources.

Google Privacy Sandbox officially shuts down: what it means and what’s next Account of the Privacy Sandbox wind-down and its practical consequences for advertising measurement and consent strategy.

Google’s antitrust ruling: what the remedies really mean for search, SEO, and AI assistants Analysis of the search remedies decision and its implications for default arrangements, data access and assistant competition.

Court issues remedies ruling in United States v. Google search case Legal summary of the district court’s remedies decision, used for the regulatory scenario rather than for any predicted outcome.

Does LLM traffic convert better than organic? A new data-backed study Site-level comparison of assistant-referred and organic search conversion rates, one of several independent datasets reporting a multiple in the four to five times range.

Google AI Overviews impact on publishers and how to adapt into 2026 Trade analysis of click-through decline for top-ranking pages on summary-bearing queries and the adaptation options available to publishers.

The secrets to organic reach in 2026 Platform-level detail on how recommendation-driven feeds allocate reach, supporting the argument that organic social behaves as leased rather than owned distribution.

Meridian: open-source marketing mix modelling Google’s open-source Bayesian marketing mix modelling framework, referenced as part of the measurement stack and as the basis for Meridian Studio and GeoX.

Robyn: an experiment-informed open source marketing mix modelling package Meta’s open-source modelling package, the other main route by which mid-sized advertisers gained access to marketing mix modelling without licence costs.

This article was prepared with the assistance of artificial intelligence tools. The content underwent expert human review, and Webiano Digital & Marketing Agency assumes editorial responsibility for its final version and publication.