A content moderation policy for an adult website is the written set of rules that says what users may not upload, how content is checked before and after it goes live, who decides borderline cases, and how users can report content and appeal decisions. For adult user-to-user platforms in Europe it must satisfy three audiences at once: Ofcom under the UK Online Safety Act, regulators under the EU Digital Services Act, and the card schemes through your acquirer. The core is a clear prohibited content list, pre-publication review, verified uploaders, fast reporting routes and records of every decision.
This how-to walks through building one, including the new UK prohibited categories that take effect on 6 November 2026. It is general information, not legal advice.
Who sets the rules your policy must meet
| Source | What it expects from moderation |
|---|---|
| UK Online Safety Act 2023 | Illegal content risk assessment; proportionate measures to prevent priority illegal content; swift takedown; reporting and complaints; a content moderation function that is resourced and trained |
| EU Digital Services Act | Clear terms on restrictions; notice-and-action; statements of reasons; internal complaint handling for online platforms; action against repeat abusers; transparency reports |
| Card-scheme rules | Verified uploaders; consent from everyone depicted; review of all content before publication; real-time monitoring of live streams; complaints resolved within seven business days; appeals for depicted people |
| Criminal law | Zero tolerance for child sexual abuse material, extreme pornography, non-consensual intimate images and, in England and Wales from 6 November 2026, three further categories |
The card-scheme standard is usually the strictest in practice, so many platforms design to it and map the regulatory duties on top.
Step 1: write the prohibited content list
Your prohibited content list is the heart of your adult content moderation guidelines. Make it specific enough that two moderators reach the same answer. Group it by severity.
Tier 1: illegal, remove and escalate
- Anyone under 18, or content suggesting a minor, real or synthetic. Escalate under your CSAM procedure; see reporting CSAM and handling law enforcement requests.
- Non-consensual content, including hidden-camera material, leaked private images and sexual deepfakes of real people.
- Extreme pornography as defined in national law.
- From 6 November 2026, for UK users: pornography depicting strangulation or suffocation, sexual activity between people presented as related (including specified step and foster relationships), and anyone presented as under 16. See Crime and Policing Act 2026 pornography offences.
- Trafficking, coercion or exploitation indicators, threats and incitement to violence.
Tier 2: breaches card rules or your terms
- Content where you cannot verify every participant's age, identity and consent.
- Impaired participants, such as visible intoxication, or content that appears non-consensual even if staged.
- Unlabelled realistic AI-generated content; see AI Act deepfake labelling.
- Copyright infringement and content uploaded by someone other than the rights holder.
- Off-platform payment solicitation, spam and impersonation.
Tier 3: restricted or allowed with conditions
- Themes allowed only with specific labels, age-gated areas or geo-restriction.
- Promotional content in public areas, which must not be explicit where it can be seen before age assurance.
Write for keywords too. Titles, tags and descriptions are content. Prohibit and filter terms that signal banned themes, and remove them from search suggestions, category pages and SEO landing pages, not just from video files.
Step 2: design the moderation workflow
A robust moderation workflow for adult uploads usually runs in this order:
- Account verification: verify the uploader's age and identity before they can post. See performer age and consent records.
- Upload declarations: the uploader names everyone who appears, attaches or links their verification and release, and declares any AI generation or manipulation.
- Automated screening: hash matching against known CSAM and non-consensual image lists, matching against your own block list, keyword filters on metadata, and classifiers to flag risk signals.
- Human pre-publication review: a trained moderator checks the content against the prohibited list and the participant records.
- Publication with labels and any geo-restrictions.
- Post-publication monitoring: user reports, periodic re-review of high-risk categories and live-stream monitoring.
- Enforcement: removal, restriction, strikes, suspension and, where needed, reporting.
Live streaming
Live content cannot be pre-reviewed in the same way. Card rules expect real-time monitoring and the ability to cut a stream. In practice that means automated alerts, a moderator rota, clear on-stream rules and recorded streams for later review where lawful.
Human vs automated moderation
| Automated | Human | |
|---|---|---|
| Strengths | Speed and scale; reliable for known content via hashes; consistent keyword filtering | Context, consent cues, age judgement, roleplay framing, appeals |
| Weaknesses | False positives and negatives; poor at new content and context | Slower; costly; welfare risks; inconsistency without training |
| Best used for | Triage, known-content blocking, prioritising queues | Final decisions on adult content, borderline cases, reports |
For adult platforms the sensible model is automation to sort and block known material, and humans to decide. Document which decisions are automated, because DSA statements of reasons must say whether automated means were used.
Step 3: escalation and specialist routes
Not every case belongs in the general queue. Set up separate, faster routes for:
- Suspected CSAM: immediate restriction, preservation and reporting by a named lead.
- Non-consensual intimate images: restrict on report, verify proportionately, hash and block; see intimate image abuse duties for sites.
- Threat to life or safety: notify authorities, as required under DSA Article 18.
- Legal and copyright notices: handled against your takedown procedure.
- Performer complaints: anyone depicted can ask for removal; if consent cannot be shown, the content comes down.
Give each route a target time. A hypothetical example: suspected CSAM restricted within one hour, intimate image reports within 24 hours and never more than 48, other illegal content reports within 24 hours, card-scheme complaints resolved within seven business days.
Step 4: notices, statements of reasons and appeals
Under the DSA, when you remove or restrict content of EU users, you generally need to give the user a statement of reasons: what was restricted, the legal or contractual ground, the facts relied on, whether automation was used and how to appeal. See our guide to DSA statements of reasons.
Online platforms must also offer an internal complaint system for at least six months after a decision, with decisions not taken solely by automated means, and must tell users about out-of-court dispute settlement. The Online Safety Act separately requires accessible complaints procedures for UK users. One well-designed appeals process can serve both.
Step 5: moderators, training and welfare
- Training: on the prohibited list, age and consent indicators, the new UK categories, and your escalation routes. Refresh it whenever the law or policy changes.
- Guidance with examples: written descriptions of borderline cases, without storing harmful material as examples.
- Quality assurance: second review of a sample of decisions and tracking of disagreement rates.
- Welfare: exposure limits, rotation, counselling access and the right to step away from a case.
- Language coverage: moderators or tools for every language you accept uploads in.
Step 6: records, metrics and review
Keep a decision log for every action: content ID, rule applied, who or what decided, timestamps, notices sent and appeal outcome. It is your evidence for Ofcom, your acquirer and DSA transparency reports. Track at least:
- volume of uploads reviewed and rejected, by reason;
- time to action for each report route;
- appeal rates and reversal rates;
- repeat offenders suspended.
Review the policy at least annually and whenever your illegal content risk assessment changes, you launch a new feature such as live streaming or AI tools, or the law changes.
How to structure the written policy
Most platforms end up with two documents: a public policy that users read, and an internal manual that moderators use. Keeping them aligned matters, because a gap between what you tell users and what you actually enforce is exactly what regulators and acquirers look for.
A practical structure for the public policy is:
- purpose and scope, including which services and features it covers;
- the prohibited and restricted content rules, in plain language;
- verification and consent requirements for uploaders and everyone shown;
- how content is reviewed, including the use of automated tools;
- how to report content, with separate routes for urgent harms;
- what happens when rules are broken, including strikes and suspension;
- how to appeal and how long decisions take;
- contact points for authorities and rights holders.
The internal manual adds decision trees, examples described in words, escalation contacts, target times and record-keeping instructions. Version both documents, date every change and keep old versions, so you can show what rule applied when a decision was made.
Frequently asked questions
Do I have to pre-review every upload?
The law does not impose a general monitoring obligation, but card-scheme rules for adult merchants require review before publication. If you take card payments, plan on reviewing everything.
Can I publish my full prohibited content list?
You should publish the rules in your terms clearly enough for users to understand them; the DSA requires clear, accessible terms on restrictions. Keep detailed moderator guidance and detection signals internal so they are harder to evade.
Do these rules apply to a creator's own site?
If only you publish content, you are a publisher, not a user-to-user service, and the duties differ. If fans can upload or post, user-to-user duties apply. See running a self-hosted creator website.
Next steps
Draft the three-tier prohibited list, map your current workflow against the seven steps above and close the gaps before 6 November 2026. For moderation tooling and outsourced moderation, browse the webmaster software and services directory, and find related guides in the laws and regulations hub.