A transparency report is a periodic public document in which an online service discloses how it moderates content, handles government and legal requests and enforces its own rules, usually with aggregated numbers for a defined reporting period. In online safety regulation, transparency reports have moved from a voluntary trust-building exercise to a legal obligation under laws such as the EU Digital Services Act and the UK Online Safety Act.
The report does not change any moderation decision by itself. Its function is accountability through measurement: regulators, researchers, journalists and users can compare what a platform says about safety with what it actually removes, restricts and reviews.
What a transparency report contains
A modern transparency report combines quantitative tables and qualitative explanations. The numbers show volumes and speed; the explanations describe policies, tools and the meaning of the categories, without which the numbers are hard to interpret.
| Component | Typical metrics | Why it matters |
|---|---|---|
| Authority orders | Number of orders by country and type, time to act | Shows state pressure on content |
| User notices | Notices received, action taken, median handling time | Measures responsiveness to reports |
| Own-initiative moderation | Removals and restrictions by policy category | Reveals the scale of proactive enforcement |
| Automated tools | Share of automated decisions, accuracy indicators | Exposes reliance on classifiers |
| Complaints and appeals | Appeals filed, decisions reversed | Indicates error rates of moderation |
| Moderation resources | Moderators by language, training | Tests capacity across markets |
| User numbers | Average monthly active recipients | Determines regulatory category |
Reports are usually published as a PDF narrative plus machine-readable data files. Comparability across platforms has long been weak, because each company defined its own categories, counted differently and chose its own periods.
Legal obligations under the Digital Services Act
The Digital Services Act makes transparency reporting mandatory for intermediary services, with obligations that grow with the size and role of the service. Micro and small enterprises are largely exempt unless designated as very large.
| Service type | Reporting frequency | Additional content |
|---|---|---|
| Intermediary services | At least once a year | Orders, notices, own-initiative moderation |
| Online platforms | At least once a year | Disputes, suspensions, active user numbers |
| Very large online platforms | Every six months | Moderator resources by language, accuracy |
| Very large online search engines | Every six months | Same as very large platforms |
An implementing regulation introduced harmonized templates and reporting periods, so reports cover fixed half-year or calendar-year windows and follow common categories. Separately, platforms must send each statement of reasons for a moderation decision to the public DSA Transparency Database, which offers decision-level data rather than aggregates.
Very large platforms also publish systemic risk assessments and independent audit reports, although with redactions. These documents explain why risks exist; the transparency report shows what was done and how often.
Other regimes and voluntary reporting
Voluntary transparency reports began around 2010, initially focused on government requests for user data and removals. They remain common for topics that law does not yet cover in detail, such as law enforcement data requests in particular jurisdictions.
| Regime | Who decides the content | Main focus |
|---|---|---|
| Voluntary corporate report | The company | Government requests, policy enforcement |
| EU Digital Services Act | Law and harmonized templates | Moderation, notices, appeals, resources |
| UK Online Safety Act | Ofcom through transparency notices | Service-specific safety information |
| EU terrorist content rules | Regulation on terrorist content online | Removal orders and actions taken |
| Australian online safety framework | eSafety Commissioner notices | Answers to targeted safety questions |
Under the UK Online Safety Act, Ofcom issues annual transparency notices to designated categories of services and specifies what each must publish, so the content of reports can vary by service type and risk. Ofcom then publishes its own summary report comparing providers.
Reading a transparency report critically
Numbers in a transparency report are easy to misread. A rise in removals can mean more harmful content, better detection, broader policy definitions or simply more users, and the report rarely isolates the cause.
- Check definitions first: a removal, a restriction and a demotion are different actions, and some reports merge them.
- Compare periods carefully: template changes can create breaks in time series that look like trends.
- Look at reversal rates: a high share of successful appeals suggests over-removal or weak accuracy.
- Watch the language coverage: moderator counts by language reveal markets with thin human review.
- Remember what is missing: prevalence of harmful content seen by users is often not reported at all.
Researchers therefore combine reports with data access for vetted researchers, audit findings and independent measurement. A transparency report is a starting point for scrutiny, not proof of safety.
Producing a compliant report
For a platform, the hard part is not writing the report but building the data pipeline behind it. Every notice, decision, appeal and automated action needs a timestamp, a policy category and an outcome that maps cleanly to the regulatory categories.
Mature teams treat reporting as a continuous compliance process: categories are fixed in the moderation tooling, counts are reconciled with the statements of reasons sent to the database, and a legal review checks consistency with risk assessments before publication. Inconsistencies between these sources are a common trigger for regulatory information requests.
Smaller services face a different problem: scale without staff. A forum or marketplace with a few employees still has to count orders, notices and moderation actions correctly once it falls under the DSA thresholds. A practical approach is to tag every decision with a fixed category at the moment it is made, so that the annual report becomes a query rather than a manual reconstruction from emails and spreadsheets.
Publication timing also matters. Under the harmonized rules, reports must appear within a set number of months after the period ends, and late or incomplete publication is itself a breach that regulators can pursue independently of any moderation failure.
In simple terms
A transparency report is a platform’s public scorecard on safety. Twice a year or once a year, the company publishes how many posts it removed, how many reports users sent, how fast it reacted and how often it got things wrong.
It works like the annual accounts of a company, but for moderation instead of money. Anyone can read it and compare promises with actual results.
A practical example
A video-sharing platform designated as a very large online platform prepares its half-year report. Its trust and safety data team exports figures from the moderation system: notices by category, decisions taken automatically and by humans, median handling times and appeal outcomes.
During reconciliation, the team finds that appeal reversals for hate speech decisions in two smaller languages are far higher than average. The report discloses the figures, and the company adds native-speaking moderators for both languages. In the next period, the reversal rate falls, which the regulator notes as evidence of a working feedback loop between reporting and moderation quality.