Age verification and GDPR have to work together: if law requires you to check users' ages, you need a lawful basis for the data involved, and you must collect only what the check needs, keep it only as long as necessary and protect it properly. In practice that means designing the check so that your site learns "over 18: yes or no" and little else, choosing a provider that deletes images and documents quickly, putting a proper contract in place and carrying out a DPIA. Regulators in the UK and EU have both said age assurance must not become a new source of tracking or profiling.

This guide covers the data protection side. Whether you have a legal duty to verify ages at all is covered in our age verification section. It is general information, not legal advice.

What regulators have said about age verification privacy

Three documents set the tone:

  • ICO age assurance opinion. The ICO updated its Commissioner's Opinion on age assurance in January 2024 to reflect the Online Safety Act, working with Ofcom to align the two regimes. It expects age assurance to be necessary and proportionate to the risks and compliant with all the data protection principles.
  • ICO and Ofcom joint working. The two regulators have published joint statements on how online safety and data protection fit together. Ofcom decides whether your check is "highly effective"; the ICO decides whether it handles personal data lawfully. You must satisfy both.
  • EDPB Statement 1/2025. Adopted in February 2025, the EDPB statement on age assurance sets out ten principles, including risk-based proportionality, preventing age assurance from enabling identification, location, profiling or tracking, strict data minimisation, data protection by design, and accountability.

Lawful basis for age checks

Which Article 6 basis applies depends on why you are checking:

SituationLikely lawful basis
UK pornography provider under Part 5 of the Online Safety Act, or an EU operator under a national law such as France's SREN ActLegal obligation (Article 6(1)(c)), where the law requires the check
Voluntary age gating where no law requires itLegitimate interests, with a balancing test, or consent
Platform verifying creators' age and identity for its own rules and to prevent illegal contentLegal obligation where applicable; otherwise legitimate interests or contract

A legal obligation covers only what the law actually requires. It does not justify reusing age-check data for marketing, analytics or fraud scoring. If you want to do any of that, you need a separate basis, and in most cases you shouldn't do it at all.

Data minimisation in age checks

This is where most of the risk sits. Good design looks like this:

  1. Your site receives only a result. The provider performs the check and returns a token or flag saying the user is over the threshold. You don't receive a date of birth, document image or selfie.
  2. The provider doesn't learn where the user is going, or learns as little as possible. Some schemes, including the "double anonymity" model promoted by France's Arcom, are built so the verifier can't see which site requested the check.
  3. Reusable tokens are unlinkable. A user who verifies once shouldn't become trackable across sites by a persistent identifier.
  4. Logs are minimal. Keep proof that a check happened (timestamp, method, result) if you need it to demonstrate compliance, not the underlying evidence.

The EDPB statement specifically points to tokenised approaches, selective disclosure and zero-knowledge proofs as privacy-preserving options, and notes that a no-log policy can be a valuable safeguard. The EU's own age-verification app is built around similar ideas; see our guide to the EU age verification app.

Biometric data and facial age estimation under GDPR

Facial images processed by specific technical means are "biometric data", but they only become special category data under Article 9 when processed for the purpose of uniquely identifying someone. That distinction matters:

  • Facial age estimation analyses a face to estimate age without checking who the person is. If the system genuinely does not identify anyone, there is a reasonable argument that it is personal data but not special category biometric data under Article 9. Check the provider's own analysis and your regulator's latest guidance. It is still high-risk processing and needs strong safeguards: immediate deletion, no reuse for training without a lawful basis, and clear information to users.
  • Document checks with a selfie match compare a live face against an ID photo to confirm it's the same person. That is uniquely identifying, so it is special category biometric data and needs an Article 9 condition.

Either way, the method's accuracy affects fairness. Estimation errors can wrongly block adults or let minors through, and they vary across age bands and demographics; see our analysis of facial age estimation accuracy.

Watch the context: the fact that someone has passed an age check for an adult site is itself sensitive. It can reveal an interest relating to their sex life. Treat check records as sensitive even when the method itself isn't biometric. See special category data on adult sites.

Retention: how long to keep age-check data

  • Images and documents: delete immediately after the check, or within the shortest window the provider can technically manage. Ask for this in writing.
  • Result tokens: keep for the session or account lifetime, as needed to avoid re-checking.
  • Audit records: keep minimal records for as long as you need them to demonstrate compliance to Ofcom or a national regulator. Set and document a period.

Processor contracts with age verification providers

Age verification providers may act as your processor (they check on your instructions and return a result) or as an independent controller (for example, operating a reusable digital ID that the user holds). Work out which, because it changes the contract. Our guide to controller vs processor roles explains the test. For a processor, the Article 28 agreement should cover:

  • the exact data processed and its deletion timeline;
  • a ban on using your users' data for the provider's own purposes, such as training models, unless agreed and lawful;
  • sub-processors and where processing takes place, including any transfers outside the UK or EEA;
  • security standards and certifications, and breach notification to you without undue delay;
  • audit rights and help with DPIAs and subject access requests.

Our guide to choosing an age verification provider covers the commercial and effectiveness questions alongside these.

Do you need a DPIA for age verification?

Almost certainly. Age assurance involves innovative technology, can block access to a service and, on adult sites, concerns sensitive contexts. The EDPB says a DPIA will often be required. Use it to compare methods and record why you chose the least intrusive option that is still effective; see when a DPIA is required.

Frequently asked questions

Is self-declaration better for privacy?

It collects less data, but Ofcom has said self-declaration isn't highly effective age assurance under the Online Safety Act. Where a law requires robust checks, you need a method that works and protects privacy.

Can I use age-check data to prevent fraud?

Not by default. That is a different purpose needing its own lawful basis and transparency, and it may conflict with what you told users. Keep the two separate.

Do affiliates need to worry about this?

If you only link to sites that do their own checks, the data protection duties sit mainly with them and their providers. If you run your own age gate, they apply to you.

Next steps

Draw the data flow for your current or planned check and ask your provider to confirm what they keep and for how long. Compare providers in our compliance and verification services directory.