Overview

Age assurance serves different purposes on social media and dating services. Social platforms generally allow children to participate, so age signals must arrive early enough to shape moderation, content, features, and recommendations. Dating services seek to enforce an 18 minimum age, yet fewer than 0.04% of users who completed checks were later identified as under 18, a metric that cannot distinguish strong exclusion from weak detection. Ofcom’s early UK evidence supports a conditional, not universal, verdict.

Learning outcomes

On social media, age assurance helps only when child results activate moderation and recommender protections early enough to matter.
On dating services, an under-18 or unresolved result is intended to stop entry, not merely tailor the experience.
A low rate of children detected after passing cannot show how many children the initial check excluded.

Keywords

Age assuranceSocial mediaDatingModerationRecommendationsExclusion

Social Media and Dating: When the Boundary Changes

This chapter compares two settings in which age assurance serves a different purpose from the front gate used by a pornography service. Social platforms generally allow children to use the service, so the key question is whether the process identifies them early enough to shape protections around content, features, and recommendations. Dating services seek to enforce a hard minimum age, so the question is whether an age decision actually keeps under-18 users out. The comparison helps the reader distinguish age-aware tailoring from complete exclusion and recognise why a low rate of later child detection is not automatically a success signal.

The same facial estimate, photo-ID match, or behavioural signal can therefore lead to different duties. On social media, a child result may activate safeguards while preserving access to ordinary parts of the service. On a dating service, an unresolved or under-18 result may be intended to end the user journey. A method cannot be judged apart from the protection state it is meant to create.

Social media: identifying a child is not the whole job

Social media services are generally designed for adults and children. Ofcom therefore describes age assurance in this setting as a way to tailor a child’s online experience rather than prevent all children from entering. That design objective changes what “effective” means. The process must identify children reliably enough for protections to apply, but those protections also depend on finding harmful content and preventing recommender systems from repeatedly presenting it.

Ofcom observed two broad approaches among the analysed services. The first was active age-gated content. A user can usually access the service without completing an age check, but must actively complete one when attempting to access a particular content category, function, channel, or community reserved for adults. Five analysed social media services used this approach. It limits the number of users who need to complete a check and can therefore reduce unnecessary collection of sensitive data. It also concentrates the safety question at the moment a user requests something the service has classified as unsuitable for children.

The method still has to be placed correctly. Content that has been designated for adults should remain behind the gate, and a user who has not completed the check and been determined to be an adult should not be able to see it. A technically accurate check that appears after the restricted content has already been shown cannot deliver the intended protection. Nor does an active gate protect against harmful content that the service has not classified or that its recommender system sends through another route.

The second approach was passive age inference. Four analysed services asked users to declare their age and then used proprietary models to infer whether the user was a child or an adult from activity, behavioural indicators, and account-level information. The models could operate on an ongoing basis rather than waiting for a visible age-check request. When confidence was low or a risk signal appeared—for example, an attempt to access age-restricted content or a change to age-related account settings—the service could trigger a more active method such as facial estimation, photo-ID matching, a credit-card check, or email-based estimation.

This approach can appear proportionate because it does not require every user to present an official identity document.

Published by DeepMox

FAQ

How does age assurance differ on social media and dating services?

On social media it generally identifies children so the service can tailor protections; on dating services it is intended to keep under-18 users out. The same method therefore leads to different access states and different consequences for error.

Why can passive age inference miss children on social media?

Passive inference can miss children because it depends on activity and account signals over time and may fail silently before a determination. Services must protect users whose status remains unresolved and show evidence about detection, timing, fairness, and robustness.

What does a low rate of later child detection on dating apps mean?

It does not show whether initial checks excluded most children, because later detection depends on reports and automated discovery after access. Fewer than 0.04% later identified as under 18 could reflect strong exclusion, weak detection, or both.

Previous10 minAge Assurance for Pornography Services

Next12 minAge Assurance: The Cost of a Check

Age Assurance on Social Media and Dating

Social Media and Dating: When the Boundary Changes

Social media: identifying a child is not the whole job

FAQ