Age inference from account signals

Existing account information, including account age or an existing credit card, may establish adulthood. Apple does not promise this route for every account.

Character.AI’s FAQ: “We’ll look at a number of signals, including login info, activity on the platform, and some signals from third parties. This is a standard way that companies do age assurance.” and “Our age assurance tech doesn’t require you to submit any extra personal data.” Its 29 October 2025 blog: “We have built an age assurance model in-house and will be combining it with leading third-party tools including Persona.” No retention wording is given for the estimate.

The age-prediction system itself. OpenAI's page: it “looks at different signals linked to your account. For example, it may look at general topics you talk about, the times of day you use ChatGPT, how and when your account is used, and how long your account has existed.” Verifying with Persona stops age prediction running on the account.

Meta’s May 2026 post gives June 2026 as the planned UK start on Facebook; no later Meta page read confirms it. Visual analysis is “currently available in select countries”.

Meta’s newsroom post of 5 May 2026 (corrected 31 July 2026): AI analyses “entire profiles for contextual clues — such as birthday celebrations or mentions of school grades” across posts, comments, bios and captions, and Meta is “continuing to expand this technology across additional parts of our apps like Instagram Reels, Instagram Live, and Facebook Groups”; “visual analysis” of photos and videos “is not facial recognition” and “does not create facial fingerprints or identify who is in a photo”. On placing suspected teens into Teen Account settings: “We’re also expanding to Facebook in the US for the first time, followed by the UK and the EU in June.” Meta’s 5 December 2022 post described “age detection technology to find potential discrepancies in the ages people provide us and the age we think they may be” for Facebook Dating. No retention statement for the signals used was found.

Google’s page: “Google may use algorithmic systems and processes to confirm your age status by reviewing existing information associated with your account. This may include: how long you’ve had your Google Account, your account activity, whether you have forms of payment in your account.” and “When you provide a birthday that indicates you’re a kid or teen, we automatically apply additional protections.” No retention wording is given for the estimate itself.

Meta: “certain advanced features — like visual analysis — are currently available in select countries as we work toward a broader rollout.”

Meta’s newsroom post of 5 May 2026 (corrected 31 July 2026): AI analyses “entire profiles for contextual clues — such as birthday celebrations or mentions of school grades” across posts, comments, bios and captions, plus “visual analysis” of photos and videos, which Meta says “is not facial recognition” and “does not create facial fingerprints or identify who is in a photo”. Accounts suspected to be teens “even if they list an adult birthday” are placed in Teen Account settings; Meta’s 21 September 2025 update said this testing expanded to the UK, Canada and Australia “starting today”, and the May 2026 post says Meta “successfully launched this technology on Instagram in the US, Australia, Canada, and the UK”. Meta says it gives people “the option to change their settings” in case of a mistake, and that an account it determines may be under 13 “will be deactivated” until proof of age is provided. No retention statement for the signals used was found.

Reddit’s page lists the signals it uses to predict an age group without asking: the account email address and domain “including email-based age inference from third-party providers”, the age of the account, posts, comments and subscriptions, the birthdate given at sign-up, prior verification status, and age signals from app stores and third-party providers. Where the system cannot determine the age group, Reddit may ask for device or partner verification.

Replika’s blog (10 April 2023): “we have a system that reacts to any mentions of being underage. If detected, we immediately prompt the user to re-enter their birthday or temporarily suspend their access to the app.” This is a reaction to what a user writes, not an estimate from behaviour; it is recorded here because Replika describes it as its means of finding under-18s. Replika’s current Privacy Policy (27 May 2026) lists “Verifying the age of registered users” as a purpose but does not describe the method. No retention period is published, so the absence line applies.

Two groups on X's own page. Signals X treats as verification without asking the user: a previously self-attested age under 18, a previously completed X ID verification, Legacy Verified status, an account created in 2012 or earlier, and grey-badged (government) or gold-badged (Premium Business) accounts. Separately, X estimates age from the phone number on the account. No retention wording is published for either.