OpenAI brings cost-per-click ads to ChatGPT and changes the search market

OpenAI has finally crossed a line it spent a long time standing near and publicly disliking. ChatGPT is no longer only a subscription product, a research assistant, or a consumer AI habit. It is now an advertising surface. The company first said in January that it would begin testing ads for some U.S. users on Free and Go, launched that test on February 9, and then expanded ad rollout in April to Australia, New Zealand, and Canada. The company’s official language has been careful from the start: ads are clearly labeled, separated from answers, kept away from sensitive topics, and absent from paid tiers such as Plus, Pro, Business, Enterprise, and Education. Reuters, AP, and OpenAI’s own support documents all point to the same story: advertising is no side experiment anymore. It is a structural revenue layer for a product used by hundreds of millions of people.

The newer twist is the one that matters most to the ad market. Trade reporting from Digiday and Search Engine Land says OpenAI has turned on cost-per-click pricing inside ChatGPT, with early CPC bids reportedly in the $3 to $5 range through a limited ads manager. That changes the meaning of the rollout. A CPM test can be dismissed as a brand pilot. A CPC system is a claim about measurable intent. It says OpenAI thinks a user chatting through a problem, comparing options, or deciding what to buy is close enough to a classic search user that advertisers will pay for action, not just visibility. That is why this shift matters beyond OpenAI itself. It pushes ChatGPT closer to Google’s commercial core, forces rivals to define their own monetization ethics, and tells publishers, merchants, and SEO teams that answer engines are no longer just about traffic loss. They are becoming marketplaces for attention, recommendation, and transaction.

ChatGPT’s ad era is no longer hypothetical

The cleanest way to understand this moment is to ignore the hype and look at the timeline. On January 16, OpenAI said it would start showing ads to some U.S. users on Free and Go “in the coming weeks,” with placements separate from answers and with no sharing of user conversations with marketers. On February 9, OpenAI published its formal launch note for the test, saying ads were beginning in the U.S. for logged-in adult users on Free and Go. On April 16, ChatGPT release notes added Australia, New Zealand, and Canada to the rollout. That sequence matters because it shows a fast transition from principle to product. This was not a distant monetization idea buried in an earnings call. It was planned, announced, deployed, and expanded inside a single quarter.

OpenAI has tried hard to frame the move as access, not extraction. Its public explanation is that ChatGPT is used by hundreds of millions of people and that keeping Free and Go fast and reliable requires heavy infrastructure investment. Reuters reported in February that ChatGPT had more than 800 million weekly active users. AP reported six days ago that OpenAI CFO Sarah Friar said the product now has more than 900 million weekly users and that about 95% of them do not pay. Once those numbers are paired with the compute bill, ads stop looking like a philosophical betrayal and start looking like revenue gravity. The company can keep pushing subscriptions and enterprise sales, and it clearly will, but a free consumer product with that scale was always heading toward some form of monetized discovery. The surprise is not that ads arrived. The surprise is how quickly OpenAI moved from “we plan to test” to “we are already expanding.”

Cost per click is the real turning point

The headline about ads launching inside ChatGPT was big enough. The shift to cost per click is bigger. Reuters reported in January that OpenAI’s initial ad offering to advertisers was based on views, not clicks, and that self-service buying technology was still in development. That looked like a cautious opening: simple inventory, limited measurement, controlled access, a new platform still learning what buyers will tolerate. By April, Digiday reported that OpenAI had turned on CPC buying, with screenshots of an ads manager showing bids between $3 and $5 per click, and Search Engine Land described the move as ChatGPT becoming a performance channel rather than only a branding environment. That is the moment when “ads in AI” turns into “AI as paid acquisition.”

Why does that distinction matter so much? Brand advertisers can tolerate fuzzy measurement on a new platform if the audience is prestigious enough. Performance advertisers usually will not. They want a cost attached to action, a model they can compare against Google Search, Meta, retail media, or affiliate traffic. Digiday’s reporting makes this plain: marketers testing ChatGPT ads can now pay when someone clicks, not merely when impressions are served, and analysts cited there describe that as the kind of consistency buyers need to justify budget reallocation. CPM says “look at our scale.” CPC says “judge us against your incumbent channels.” OpenAI would not make that move unless it believed conversation in ChatGPT can express commercial intent strongly enough to survive that comparison. It may not win that contest yet, but pricing is often the clearest statement of strategic ambition.

Conversation is now inventory

The most important sentence on OpenAI’s advertiser page is not the sales pitch. It is the description of user behavior. OpenAI says people use ChatGPT to ask quick questions, explore ideas, compare options, and plan what to do next, and that as those conversations evolve, people naturally move from exploring to making decisions. That is the commercial thesis in one paragraph. OpenAI is not selling pageviews. It is selling the moment between uncertainty and action. A classic display ad interrupts content. A search ad intercepts intent. A ChatGPT ad tries to appear during intent formation itself, while the user is still describing the problem in natural language and before the shortlist has fully hardened.

That creates a different kind of advertising surface. Search keywords often flatten meaning into a terse string. Conversation carries more texture: the constraint, the budget, the anxiety, the trade-off, the style preference, the timing problem. OpenAI’s own help documentation says ad matching starts with what is being discussed in the current thread and, if a user chooses personalized ads, can also draw on past chats, memory, and ad interaction history. Google now does something related in AI Overviews, matching ads to both the user query and the content of the generated overview. The difference is tone and intimacy. Google still feels like search with AI layers. ChatGPT often feels like delegated thinking. That is why the ad opportunity is so attractive and why the trust risk is sharper. A user does not feel “served” when talking to ChatGPT. The user often feels understood. That feeling is commercially potent and politically dangerous at the same time.

The answer and the ad are being kept deliberately apart

OpenAI knows the trust problem, which is why nearly every official explanation repeats the same architecture. Ads do not influence answers. Ads run on separate systems from the chat model. Advertisers cannot shape, rank, or alter responses. Ads appear below the end of a response, are clearly labeled as sponsored, and are visually separated from the organic answer. The company repeats the point across its launch post, its help center, and its broader principles document. This is not just a product choice. It is the legal and reputational foundation of the whole business. If users start to believe answers are quietly bent toward paying partners, the ad product damages the thing it depends on.

OpenAI has also been careful to distinguish ads from other commercial surfaces in ChatGPT. The shopping help page says product carousels shown for shopping intent are selected independently by ChatGPT, are not ads, and are not influenced by OpenAI partnerships. The “Buy it in ChatGPT” launch post makes the same claim about organic product results being ranked on relevance to the user. That separation matters because commerce inside ChatGPT now has multiple layers. There are organic product recommendations, structured merchant feeds, shopping research guides, instant checkout, and now sponsored placements. If OpenAI let those boundaries blur, every product answer would start to feel suspect. It is trying to build a world in which a user can trust the recommendation and still accept a sponsored next step nearby. That is a hard balance. It may be the hardest product problem in consumer AI.

Privacy promises are doing as much work as the ad tech

A lot of the early coverage around ChatGPT ads focused on the presence of ads. The deeper issue is the data model behind them. OpenAI says conversations are kept private from advertisers, user data is never sold to advertisers, and advertisers receive only aggregated, non-identifying performance information such as total views or clicks. The consumer data FAQ says OpenAI does not share chat content for marketing or advertising purposes. Those are strong claims, and OpenAI keeps repeating them because it knows it is asking users to accept a commercial layer inside a product that often handles work questions, relationship stress, study problems, shopping plans, health anxieties, and everyday confusion. The company is not just selling ad inventory. It is asking for continued emotional permission to exist in personal contexts.

The tension sits in the personalization details. OpenAI says ads are matched to the current chat thread. If ad personalization is turned on, it may also use past chats, memory, and how the user has interacted with ads. It also says ads can use general context such as language or broad location while never passing chats, chat history, memories, precise location, name, email, or IP address to advertisers. Users can turn personalization off, clear ads data, and use Temporary Chat to avoid using or updating memory. That is a meaningful control set, but it does not erase the structural fact that conversation itself has become targeting context inside the product. OpenAI is right to draw a line between internal relevance systems and advertiser access. Users may still ask a simpler question: if the system needs to understand my conversation well enough to commercialize it, how different does that feel from being profiled, even if the profile never leaves the platform? That question will not disappear through copywriting.

The Free plan now has three prices

One of the smartest and least discussed details in OpenAI’s ads rollout is that the Free plan no longer has a single shape. The help center says Free users in test regions can continue using the ad-supported version, switch to an Ads-Free experience with reduced usage limits and fewer features, or upgrade to Plus or Pro for a broader ad-free experience. In other words, OpenAI has created a three-part monetization ladder inside what used to look like one product tier: pay with attention, pay with limitations, or pay with money. That is more sophisticated than a simple “ads for free, subscriptions for paid” split. It gives OpenAI a way to learn what people value more: feature depth, uninterrupted experience, or zero monthly cost.

That design also says something important about OpenAI’s strategic confidence. A company worried only about ad fill would force ads on free users and call it done. OpenAI is doing the opposite. It is keeping paid tiers clean, offering a limited ad-free Free path, and publicly promising that long-term value matters more than maximizing time spent. That does not make the company altruistic. It shows the product team understands that AI assistants are not social feeds. A feed can often survive a mediocre ad experience. A tool people use for writing, planning, coding, and personal problem-solving is more fragile. By letting users escape ads either through subscription or through reduced limits, OpenAI is effectively saying that choice itself is part of the trust bargain. It is also collecting market data on willingness to trade friction for privacy-like comfort. That is not a side detail. It is core product economics.

Shopping inside chat was already moving toward commerce

The ad rollout did not happen in isolation. OpenAI had already been building shopping and commerce features that moved ChatGPT closer to a decision engine. In late 2025, the company launched shopping research, describing it as a deeper buying guide that asks clarifying questions, researches across the internet, and uses memory when available to tailor recommendations. Soon after, OpenAI launched instant checkout with the Agentic Commerce Protocol, letting some U.S. users buy directly from certain merchants in chat. In March 2026, OpenAI pushed further with “Powering Product Discovery in ChatGPT,” tying richer shopping experiences to merchant feeds and promotions. Ads are arriving inside a product that was already teaching users to shop conversationally. That is why the CPC move feels less like a sudden turn and more like the monetization layer of an existing product direction.

OpenAI’s commerce infrastructure makes the SEO and retail implications more serious than a standard ad launch. The Agentic Commerce Protocol is described by OpenAI as the connective layer between merchants and shoppers in ChatGPT, and the product feed spec says merchants can upload structured catalog data so ChatGPT accurately indexes and displays products with current price and availability. That tells merchants what the new game looks like. Visibility in ChatGPT is no longer only an editorial or content problem. It is increasingly a data supply problem. If your catalog is poorly structured, your prices stale, your product attributes thin, or your seller context weak, you are not merely losing search rankings. You are becoming less legible to an answer engine that may soon decide both organic visibility and paid opportunity in the same interface. OpenAI has not built a classic search engine clone. It is building a commerce-aware answer layer with ads attached.

Brands are not buying keywords in the old sense

A lot of marketers will look at ChatGPT CPCs and instinctively translate them into paid search language. That helps up to a point and then starts to mislead. OpenAI’s help center says ad selection starts with the current thread. With personalization on, it can also use past chats, memory, and ad interactions. The ranking rule, at least in the official description, is simple: if multiple advertisers are eligible, OpenAI shows the one most relevant to the chat first. That sounds less like keyword bidding in the old search sense and more like semantic auction logic over live conversational context. Brands are not buying the query string alone. They are buying proximity to a line of reasoning.

The early category rules show how controlled that market still is. OpenAI’s ad policies say the launch phase is focused mainly on consumer categories such as lifestyle and household goods, local services, travel and entertainment, and digital products and education. The same policy blocks or sharply restricts political content, healthcare, dating and sexual content, alcohol, drugs, gambling, many financial and legal services, and other sensitive topics. Ads are also barred from vulnerable user-model interactions and from conversations around mental health, self-harm, political content, and regulated advice. That tells you the initial inventory pool is intentionally narrow. OpenAI is not trying to maximize short-term ad demand by opening the floodgates. It is trying to prove that commercial relevance can exist inside ChatGPT without poisoning the product. That is a sensible launch philosophy, though it also means early performance data will come from a limited slice of the economy.

A quick map of the new AI ad market

The easiest mistake in this discussion is to treat “AI ads” as one thing. The products already differ in placement, targeting logic, commercial posture, and trust promise.

Three models shaping AI discovery

This table matters because the market is splitting by product identity, not just by ad format. ChatGPT is building a conversational commercial layer, Google is extending search advertising into AI summaries, and Anthropic is trying to turn ad refusal into a trust signal. That leaves marketers, merchants, publishers, and users dealing with three very different ideas of what an AI assistant should be.

Google already proved the category would exist

OpenAI is not inventing the idea of ads adjacent to generative answers. Google got there first, and the official documentation is blunt. Google says ads are eligible to appear above, below, or within AI Overviews, that the system considers both the user query and the content of the AI Overview when serving ads, and that existing Search, Shopping, Performance Max, and related campaigns can participate. Google also says advertisers cannot directly target only AI Overview placements, cannot opt out of serving within AI Overviews, and do not yet get segmented reporting specifically for those placements. That is important because it shows the commercial grammar of AI answers was already being written inside the search market’s dominant ad machine.

The deeper parallel is not merely placement. It is intent expansion. Google’s own help example shows an informational question about a green pool leading to an AI Overview and then to relevant commercial ads for pool vacuum cleaners because the system infers a next step. OpenAI is now making a similar bet inside ChatGPT: conversational context can expose actionability before the user writes an obviously commercial query. The difference is starting point. Google begins with search and stretches into richer reasoning. ChatGPT begins with dialogue and stretches into commerce. That is why the OpenAI move matters to Google even if ad dollars do not shift overnight. If enough users start doing product comparison, trip planning, software evaluation, or purchase research in chat first, then the old handoff from “question” to “search results” starts to weaken. Google can defend that territory because it already has the ad system. OpenAI is trying to attack it by changing the interface where intent gets formed.

Anthropic is selling the opposite promise

OpenAI’s ad rollout has not only created a new product category. It has handed rivals a clean marketing contrast. Anthropic used it immediately. In February, the company published “Claude is a space to think,” stating that including ads in conversation would be incompatible with what Claude is meant to be and that Claude would remain ad-free. The wording is unambiguous: users will not see sponsored links adjacent to Claude conversations, and responses will not be influenced by advertisers or contain third-party product placements users did not ask for. That is not just a product policy. It is positioning. Anthropic is telling users that a serious assistant should not have mixed motives.

This contrast matters because OpenAI itself understands the argument. Reuters reported analyst warnings in January that clumsy or opportunistic ads could push people toward rivals like Gemini or Claude. AP quoted critics from the Center for Democracy and Technology arguing that users rely on chatbots as companions and advisors, and that personalized ads inside that environment could exploit trust in ways social platforms already taught the public to fear. Once ads enter the room, the product stops being judged only on accuracy, speed, and usefulness. It is judged on motive. Anthropic is betting that some users, especially high-value professional and reflective users, will pay for motive clarity. OpenAI is betting that billions of people want strong AI access without subscription costs and will accept commercial sponsorship if the boundaries stay legible. Both bets are rational. Only one can define the mainstream category.

Publishers should treat this as a distribution shock

For publishers, the ChatGPT ad story is not mainly about who gets to buy placements. It is about who gets displaced from the path between question and click. A user asking ChatGPT to compare running shoes, plan a trip, choose accounting software, or find a local service used to generate traffic somewhere across search results, affiliate content, comparison sites, or publisher guides. Now the same user can get a synthesized answer, maybe a shopping carousel, maybe a sponsored placement, maybe a path to checkout, all without the old click chain. Google’s AI Overviews already pushed this direction by putting generative summaries and ad inventory around the query layer. ChatGPT extends it into a deeper mode of task resolution.

That does not mean publishers disappear. It means their role changes. OpenAI’s shopping research product explicitly says it draws from high-quality sources across the internet. Google says AI Overviews can send traffic to a greater diversity of websites and claims the clicks from those links are higher quality. Those statements should be read carefully, not romantically. The platforms still need web sources. They still need product data, reviews, reporting, and expert content. But they now mediate value more aggressively before the click happens. For publishers, that makes citation-worthiness, entity clarity, and source authority more important, not less. It also makes pure “top of funnel” traffic models shakier. If the answer engine resolves more of the user’s uncertainty and then sells the next step itself, the publisher can still inform the answer while losing the action. That is a meaningful redistribution of value.

SEO and GEO are moving closer to merchant operations

A lot of SEO commentary still treats AI discovery as a content formatting problem. That is already too narrow. OpenAI’s shopping help page says product options may be shown when a question suggests shopping intent. The product feed specification says merchants can provide structured files so ChatGPT can index products accurately with current price and availability. The Agentic Commerce Protocol defines itself as infrastructure between merchants and shoppers in ChatGPT. That is not a publishing story alone. It is an operations story. Feed quality, schema completeness, pricing freshness, image quality, inventory accuracy, and seller context are becoming first-order visibility variables.

That shift is where SEO, GEO, and commerce start to fuse. If you want to be discoverable in AI interfaces, you still need strong editorial signals: trustworthy content, clean entities, useful comparison pages, expert language, consistent brand associations. But merchants now also need to think like platform integrators. Your content has to be citable and your catalog has to be machine-readable. OpenAI says merchant feeds can flow through third-party providers such as Salesforce and Stripe and will support future experiences such as personalization, local availability, and ETAs. That line should get the attention of every commerce team. It suggests ChatGPT is becoming not just a recommender but a fulfillment-aware environment. In that world, classic SEO work without feed discipline starts to look incomplete. The winning brands will not merely “rank.” They will be structurally legible to AI systems across content, commerce data, and post-click reliability.

Measurement will decide whether marketers stay

Early excitement does not build durable ad channels. Measurement does. Reuters reported in January that OpenAI did not yet offer easy self-service ad buying and was working to get those systems running. By late April, trade reporting described a limited ads manager and a shift to CPC bidding. That is progress, but it still looks like a platform in assembly. OpenAI’s own help materials say advertisers receive aggregated performance information such as total views or clicks, not chat content or personal details. Google, for its part, admits it does not yet offer segmented reporting specific to AI Overview placements. The common thread across AI ad products is that the user experience moved faster than the measurement stack.

CPC helps because it gives buyers a familiar yardstick, but it does not solve the whole problem. Marketers still want to know which conversations produce qualified traffic, what creative language holds up inside chat, how landing pages should be adapted for users arriving from synthesized recommendations, whether conversion intent from AI is broader or narrower than search, and how attribution should work when the assistant has already compressed research into one interaction. Those are not technical footnotes. They will decide whether ChatGPT becomes a meaningful line item or a curiosity budget. The likely path is the same one every new performance surface follows: advertisers will accept blurry reporting only if the audience is unusually high-intent or unusually hard to reach elsewhere. OpenAI’s bet is that both are true. The market has not fully decided yet.

Brand safety is not a side issue

OpenAI’s ad policies read less like an expansion playbook than a containment plan. The company bars ads near sensitive user contexts, including mental health conversations, vulnerable interactions, regulated advice, political content, weapons, self-harm, and many other categories. It also restricts or excludes many advertiser verticals at launch, including political content, healthcare claims, legal services, gambling, alcohol, adult content, and much of financial services. Reuters reported that OpenAI would not show ads to users under 18 and planned to block ads around health and politics from the start. This is not just compliance caution. It is an attempt to preserve the product’s moral center while introducing commerce.

That caution will shape the economics. Narrower categories mean less inventory and slower scale. They also mean better odds that early users see ads for ordinary consumer decisions rather than highly sensitive persuasion. OpenAI is trying to start in the most defensible commercial zones: home goods, local services, travel, digital products, education. That is smart. It also reveals what kind of ad market the company is building first. This is not an everything exchange yet. It is a controlled commerce corridor. If the product earns trust there, categories will widen. If it stumbles, critics will argue that even the cleanest categories were too close to the answer surface. Either way, brand safety is not a support function bolted onto the product later. It is one of the product’s core design principles.

The regulatory and ethical arguments are only getting started

OpenAI’s safeguards are serious, but they do not end the ethical debate. AP quoted Miranda Bogen of the Center for Democracy and Technology warning that personalized ads in chatbots could take OpenAI down the same risky path social platforms took years ago. AP also captured the intuitive objection in plain terms: people use chatbots as companions and advisors, and there is something uniquely unsettling about turning that trust into a commercial channel. The concern is not simply privacy leakage. It is role confusion. A search engine, a feed, a map, and a digital assistant each carry different expectations. ChatGPT often lands closer to “advisor” than to “media property,” and regulators eventually notice when products violate the social meaning users assign to them.

There is another regulatory wrinkle here. OpenAI says advertisers never see chats, chat history, memories, or personal details, and that sensitive information is not shared with them. That is a strong defense against the ugliest version of ad targeting. But the product still uses conversational context internally to select relevant ads. That raises a subtler policy question: when a system monetizes intimate context without exporting it, what rules should govern that internal inference layer? Existing privacy law often focuses on collection, transfer, and sale. AI assistants complicate the picture because the value extraction can happen inside the same service boundary. The system may not need to sell your data to commercialize what it knows. Law has been slower than product design on this point. That gap will not last forever, especially if AI assistants keep moving closer to health, finance, education, and work.

Users will tell OpenAI whether chat can carry commercial intent

The product question beneath all the market noise is simple. Do people experience a helpful sponsored suggestion in ChatGPT as more like search, more like shopping assistance, or more like intrusion? OpenAI is doing everything it can to nudge the answer toward the first two. Ads are below the response. Paid tiers remain clean. Sensitive topics are excluded. Personalization can be turned off. Users can ask why they are seeing an ad, hide it, report it, and clear data used for ads. Those are not decorative controls. They are the company’s way of testing the social tolerance of this format.

The company may also have timed this well. Plenty of users already treat ChatGPT as a place to compare products, plan purchases, discover tools, and ask for recommendations. OpenAI’s own shopping products were built around that behavior, and the advertiser page explicitly targets moments when users are researching and ready to act. If the sponsored placements stay relevant and low-friction, many people will accept them the way they accept promoted results in search or sponsored products in retail marketplaces. But acceptance is not the same as trust. Users may tolerate a commercial layer while quietly deciding that the paid version, an ads-free reduced Free mode, or a rival assistant feels cleaner. That is why retention behavior will matter as much as click-through rate. The first ad metric tells OpenAI what advertisers think. The second tells it whether the product meaning is changing under user feet.

The bigger shift has less to do with ads than with intent

It is tempting to frame this entire development as OpenAI borrowing a business model from Google. That is true at one level and far too shallow at another. The real shift is that ChatGPT is becoming a place where intent does not just get expressed; it gets formed, refined, and acted on. Ads are one monetization response to that fact. Merchant feeds are another. Shopping research is another. Instant checkout is another. The product is drifting toward a world where the assistant mediates the full chain from question to consideration to comparison to purchase. Once that chain lives inside one interface, the old boundaries between SEO, paid search, affiliate discovery, product recommendation, and checkout start to blur.

That is why the CPC rollout matters even to people who will never buy a ChatGPT ad. It is the clearest proof so far that OpenAI believes the assistant has become commercial territory, not just utility software. It also forces a harder question about the future of the web. If answer engines increasingly mediate discovery, keep the user longer, and sell the next step themselves, who still owns the margin created by human curiosity? Google has been asking that question in search for years. OpenAI is now asking it in conversation. The companies will compete over budgets, yes. They will also compete over something larger: the right to sit closest to human intention at the moment it becomes economically useful. That contest will shape far more than ad pricing. It will shape who gets found, who gets trusted, who gets clicked, and who gets paid.

Common questions about ChatGPT ads

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

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

Testing ads in ChatGPT
OpenAI’s official launch note for the U.S. ad test, including plan eligibility, labeling, and answer-independence claims.

Our approach to advertising and expanding access to ChatGPT
OpenAI’s principles document explaining mission alignment, privacy, control, and the initial ad rollout design.

Ads in ChatGPT
OpenAI Help Center documentation covering ad placement, personalization controls, data handling, excluded topics, and ad-free options.

ChatGPT — Release Notes
OpenAI’s release log confirming ad rollout expansion to Australia, New Zealand, and Canada.

Ad policies
OpenAI’s advertising policy framework, including allowed categories, disallowed verticals, and sensitive-context restrictions.

Advertise with ChatGPT
OpenAI’s advertiser-facing page describing the commercial thesis behind advertising in ChatGPT.

Shopping with ChatGPT Search
Help documentation distinguishing organic shopping results from ads and explaining product selection logic.

Powering Product Discovery in ChatGPT
OpenAI’s announcement about richer shopping experiences, merchant feeds, and the commerce infrastructure behind them.

Introducing shopping research in ChatGPT
OpenAI’s explanation of its buyer-guide style shopping experience and its use of conversational context.

Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol
OpenAI’s launch post for in-chat purchasing and organic product results inside ChatGPT.

Agentic Commerce Protocol
Developer documentation describing ACP as the connective layer between merchants and shoppers in ChatGPT.

Products – Agentic Commerce
OpenAI’s product feed specification showing how merchants make catalog data legible to ChatGPT.

Data Usage for Consumer Services FAQ
OpenAI’s consumer data FAQ, including statements on chat content and advertising-related sharing.

OpenAI to test ads in ChatGPT in bid to boost revenue
Reuters reporting on OpenAI’s January announcement, including early positioning around revenue pressure and trust risks.

OpenAI to start offering chatbot ads to advertisers, The Information reports
Reuters coverage of OpenAI’s early advertiser rollout, view-based pricing, and self-service tooling in development.

OpenAI CEO says ChatGPT back to over 10% monthly growth, CNBC reports
Reuters report used for ChatGPT scale context, including the more than 800 million weekly active user figure.

OpenAI rolls out age prediction on ChatGPT
Reuters coverage of OpenAI’s age-prediction system, relevant to under-18 ad exclusions and safety controls.

OpenAI plans to introduce ads for ChatGPT
AP reporting on launch details, the trust debate, and external criticism of personalized ads in chatbot environments.

ChatGPT maker OpenAI shifts its focus to business users amid Anthropic pressure
AP reporting used for OpenAI’s newer user-scale and monetization context, including the share of non-paying users.

OpenAI turns on cost-per-click ads inside ChatGPT
Digiday’s reporting on CPC activation, reported bid ranges, and advertiser interpretation of the shift.

OpenAI adds CPC ads to ChatGPT
Search Engine Land’s coverage framing ChatGPT ads as a new performance channel competing for search-style budgets.

About ads and AI Overviews
Google Ads Help documentation explaining how ads work above, below, and within AI Overviews.

Innovations in generative AI and marketing
Google’s marketing blog post on testing ads in AI Overviews and the broader commercial logic of AI-powered search experiences.

Claude is a space to think
Anthropic’s official statement that Claude will remain ad-free and why it sees advertising as incompatible with its product philosophy.