Age assurance in dating apps: What actually clears the bar

Age assurance became law in the UK, the EU, and a growing list of US states. It reaches your users as something else: a guarantee about who is allowed on the platform. Build it for the auditor and you lose the part users feel.

That part is not abstract. Users who meet a fake profile or a scam rarely file a complaint. They leave, and the churn shows up in your dashboard with no label on it.

What regulators actually reject

Ofcom’s enforcement record is a better guide to the standard than the statute is. When it fined AVS Group £1m in December 2025, the failure was specific: the service asked users to upload a photo of an ID, with no liveness detection. Ofcom held that a check a child can pass with a borrowed picture is not capable of being highly effective.

Read that as a product requirement rather than a legal one. A document check with no live face attached to it does not clear the bar. Age assurance fines so far have ranged from £450,000 to £1.35m and have all landed on services hosting pornographic content, so the amounts matter less than the standard. Dating services sit inside the same duty.

Three things teams get wrong

“Age estimation isn’t accurate enough at 17 versus 18.” True, and largely irrelevant to your funnel. Roughly 80% of the EU population and 78% of the US population is 18 or over, and none of the remainder is a small child opening a dating app. The group that genuinely needs a step-up check is much smaller than the argument about the 17-year-old edge case implies. Accuracy at that margin still matters. It just does not set your conversion rate.

“We need to know the user’s age.” You need to know they are eligible. Confirming “adult: yes” without retaining the document or the biometric template is simpler to build, easier to defend under GDPR, and easier to explain to someone already uneasy about handing over an ID. Even where a document check is required, nothing obliges you to keep the document once the decision is made.

“Estimation and verification are the same product.” A selfie-based estimate and a government ID check have different accuracy and completely different friction profiles. Defaulting to the document for everyone means paying for assurance you do not need on the large majority of signups.

The architecture that follows

A risk-based waterfall, three tiers deep.

Most users pass a selfie check in seconds. Veriff’s Age Estimation confirms eligibility without a document and without interrupting onboarding. Borderline estimates and risk-flagged sessions escalate to Identity and Document Verification, covering 12,500+ document specimens across 230+ countries and territories. The narrow top tier, suspected synthetic identities and repeated account cycling, gets biometric matching and liveness, because a genuine document proves nothing about who is holding it.

Legitimate users meet the least friction. Determined bad actors meet more of it at every layer.

Three questions for a vendor

Does the age signal reach the fraud stack? A profile caught at the age gate should change how that identity is treated for the rest of its life on the platform. Fragmented tools produce fragmented signals.

Show me the audit record. Regulators want evidence per decision, not a certificate on a slide.

Show me the conversion data, not the accuracy benchmark. Verification friction is a CAC line item. If the Trust and Safety investment is visibly draining the acquisition funnel, the trade has already gone negative.

Trust is the product

Build trust in your dating app from the first match

Discover how identity verification reduces fraud, boosts user confidence, and improves retention in online dating platforms.

Get the report