Character AI Age Verification: Is Character AI Safe?

Character AI's age verification is real, it applies to everyone, and it is not going away. An in house age estimation model scores your account first, and if it flags you as possibly under 18 the platform escalates to Persona, a third party identity service, which asks for a government ID or a face scan. The company itself is legitimate and now spends heavily on safety. It is still not a private service, and if you are in the United States you cannot opt out of your conversations being used to train its models.

That last sentence is the part almost no coverage mentions, and it is the one that should decide whether you keep using the platform. Through April 2026 the checks reached accounts that had been active for years, locking their chats to read only until the user verified, and a help page update on 14 July 2026 then narrowed that read only state to accounts verified as under 18. Below is what has actually been documented, with dates and primary sources, separated cleanly from the speculation that fills most search results on this topic.

How does Character AI age verification work?

It runs in two stages. Character.AI said in its own announcement that it was rolling out "an age assurance model in-house" and "combining it with leading third-party tools including Persona." The in house model reads behavioral signals: how you write, what you search for, your account history, the platform's own read of your login patterns. Most accounts never see anything beyond this stage.

If that model is not confident you are an adult, your account gets restricted and the only route back is the Persona check. Persona is an established identity verification vendor used across fintech and marketplaces, and the check takes one of two forms: a photo of a government issued ID, or a selfie based facial age estimate. You are not choosing between them for privacy reasons so much as for whichever one the flow offers you.

The important detail for adults is that this is probabilistic. A model guessing your age from writing style will misclassify people, and the ones it misclassifies most are adults who write casually, adults who mostly do roleplay, and accounts created years ago when the user genuinely was younger. Being asked to verify is not an accusation. It is a classifier being uncertain.

How do I verify my age on Character AI?

Go to Settings, then Advanced, then Verify Age. That path is the same on the mobile app and on the web, and it is the answer the help center gives verbatim: "On the mobile app: Go to Settings -> Advanced -> Verify Age" and "On the web: Go to character.ai, then go to Settings -> Advanced -> Verify Age." From there Persona takes you down one of two paths: a selfie that estimates your age from your face, or a photo of a government issued ID. Most people only ever see the selfie, and clearing it restores full access straight away.

You do not have to wait for the pop-up. Character.AI says age assurance "is being introduced step by step, so the pop-up might not show up for everyone right away," and that if you never saw it, the likely reasons are "your app is out of date or the prompt was previously dismissed." The Settings path works whether or not the prompt ever appeared. If the option is missing on mobile, update the app first, then restart it.

You can also switch devices partway through. The help center says verification "can be completed on any device," so if your laptop webcam is poor you can stay logged in, scan the QR code the flow shows you, and finish the selfie on a recent phone. That single step fixes a large share of the failures people blame on the check itself.

A few things make it go smoothly. Do it in daylight or under even indoor light, against a plain wall, with no hat, no sunglasses and nothing covering your jaw. Hold the phone at eye level rather than below your chin, because a low angle is one of the things that makes an adult read as younger. If you are offered the document route, photograph the ID flat on a dark surface with no flash, since glare across a hologram is the single most common reason a scan gets rejected.

Do the whole thing in the app. Verification prompts live inside the product, so any request to verify that reaches you by email or direct message is a phishing attempt, and an ID is exactly what that kind of message is fishing for.

What if Character AI age verification is not working?

Three failures account for almost all of it: a selfie the model cannot score confidently, a document photo it cannot read, and a loop that drops you back at the prompt after you thought you had passed. Lighting causes most of them. Retry in daylight against a plain background, in the app rather than a mobile browser, and the pass rate improves noticeably.

If the selfie keeps failing, take the ID route instead rather than attempting the face scan a fifth time. Repeated attempts do not train the system to agree with you, and each one is scored independently. If the ID also fails, the usual culprits are a cropped edge, a cover over the card, or an expired document. If you are stuck in the loop after a successful check, sign out fully and back in, since the status change does not always propagate to an already open session.

Nothing about a failure is permanent or destructive. The account is restricted, not deleted, and your characters and conversation history stay where they are. There is no countdown forcing you to resolve it, which means you can leave it alone and decide later.

Character.AI's own troubleshooting page names the specific things that get a selfie declined, and two of them surprise people. The first: "if the system detects added features such as drawn-on facial features, filters, makeup effects, or other changes that make you appear older or different than you are, your verification may be declined." A beauty filter you forgot was switched on will fail you. The second: "make sure you're the only person in the frame. Verification can fail if more than one face is detected," which quietly explains a lot of failures in shared rooms.

If you get routed to the document step, the rule that trips most people is that the picture has to be of the card itself. The help center is explicit that "the image must be of a physical ID, not a digital or electronic copy," so a screenshot of a mobile driver's license will not clear. Eight documents are accepted: driver's license, certificate of citizenship, consular ID, identification card, passport, passport card, permanent residence card, and immigration visa.

One line on that page settles the question people keep asking in forums. Character.AI writes: "Please note there is no alternative to submitting a selfie or government ID if you are prompted to do so. These are the only accepted methods for completing age verification." There is no email appeal, no support override, and no third route.

Why is Character AI age verification so frustrating?

Because the help pages are not resolving people's problems, and Character.AI's own site measures that. Every Zendesk help article records whether readers voted it helpful, and those counts are public. Across the eight age assurance articles, checked on 16 September 2026, 10,708 of 14,284 votes marked the page unhelpful. That is 75 percent, and it is the company's own telemetry rather than anyone's opinion.

The two worst scores are the two pages people reach when something has already gone wrong, which is the pattern you would expect if the documentation explains the policy well and the failure modes badly. For context, the oldest article in the set, "Why is Character.AI asking about my birthday?", has drawn 40,433 votes since October 2024, 20,803 of them unhelpful. It is the most voted page in the entire help center by a wide margin.

Two things follow from this that are worth acting on. If a help page did not solve your problem, you are in the large majority and you have not missed an obvious step. And all eight of those articles were edited between 14 and 16 September 2026, so the rules are still moving. Anything you read about this topic that was written a few months ago is describing a different system.

Can you bypass Character AI age verification?

No, and none of the workarounds circulating in 2026 survive contact with the system. A VPN does nothing because the check is not geographic. Borrowed or edited photos fail liveness detection, which is the specific thing Persona is built to catch. Making a fresh account restarts the same behavioral scoring and costs you the history you were trying to protect.

It is also worth being straight about the stakes. Attempting to defeat an age check is a terms of service violation, and unlike a failed selfie, which just leaves you restricted, a detected circumvention attempt can cost the account outright. The upside is a chat product. The downside is losing years of conversation permanently. That is a poor trade even when the workaround looks like it might hold.

If you are an adult who simply does not want to hand over biometric data, the honest options are the ones nobody frames as a hack: keep the reduced account, which persists indefinitely and costs nothing, or move to a platform that does not run document checks. Several do not. We compared which ones ask an adult for ID and which do not in our guide to the best AI companion app without ID verification.

Did Character AI remove age verification?

No. It is still running, and it has been expanding rather than winding down. All eight of the age assurance help articles were updated between 14 and 16 September 2026, the verification flow is still reachable from Settings, and nothing has been announced about retiring it. The rumor recurs because the rollout arrived in waves, so people who were never flagged assume it was switched off while accounts around them were being checked.

What did change is who stays restricted. Through April 2026 the checks reached long standing accounts and locked their chats to read only until the user verified. A help page update on 14 July 2026 narrowed that read only state to accounts verified as under 18, so adults who clear the check keep full access. That is the change people half remember as a removal.

Can you use Character AI without age verification?

Yes, most people can, because the check is not applied to everyone. Character.AI's FAQ is direct about it: "We'll run our age assurance technology on all accounts. But as we said above, you won't be asked to verify your age, i.e. you won't be asked for a selfie, unless we believe you're under 18 and you want to prove that you're actually 18+." If the behavioral model reads you as an adult, you will never see a selfie prompt at all.

If you are flagged and decline, you keep a reduced account for as long as you like. Nothing is deleted, there is no deadline, and your characters and chat history stay put. What you lose is open ended chat, which for most people is the entire reason they were there. So the real choice is between verifying, accepting a browsing only account, or using a platform that does not run document checks on adults. Plenty do not, and we compared which ones ask for what in our guide to the best AI companion app without ID verification.

The signals that decide this are not published in detail. The company says it looks at "a number of signals, including login info, activity on the platform, and some signals from third parties." An account created years ago by someone who genuinely was younger then, or an account used mostly for roleplay, is likelier to trip the model, which is why so many adults with long histories were the ones caught.

Why did Character AI add age verification?

Litigation and a new state law, in that order.

On 29 October 2025 Character.AI announced it would remove open ended chat for under 18 users, writing that "we will be removing the ability for users under 18 to engage in open-ended chat with AI on our platform," effective no later than 25 November. During the transition it capped teen accounts at two hours a day. It also announced it would establish and fund an independent nonprofit AI Safety Lab. Age assurance exists to enforce that ban, and a ban you cannot enforce is a press release.

The pressure behind it was concrete. In January 2026, court filings reported by CNBC, Axios, the Washington Post and CBS News showed Google and Character.AI had agreed to settle five lawsuits brought by families in Florida, New York, Colorado and Texas, including the case brought by Megan Garcia after the death of her 14 year old son Sewell Setzer III. Settlement terms were not disclosed. Those were the first resolutions in the wave of suits over chatbots and teen self harm. In May 2026 Pennsylvania sued the company separately, alleging a chatbot presented itself as a medical professional.

Running alongside that, California's SB 243 took effect on 1 January 2026. It applies to any operator making a companion chatbot available to a Californian regardless of where the company is based, and it requires clear disclosure that the bot is not human, crisis protocols that route self harm disclosures to real crisis services, recurring break reminders for minors, and annual reporting to the state beginning 1 July 2027. It also created a private right of action: $1,000 per violation or actual damages, whichever is greater, plus fees.

Is it safe to give Character AI your ID?

Your ID does not go to Character.AI. It goes to Persona, which is a specialist vendor whose entire business depends on not leaking identity documents, and which handles this for regulated financial companies. In the narrow sense the question is usually asked, yes, this is a more careful custodian than the average app you have uploaded a document to.

There is one published commitment worth knowing before you decide, because it changes the shape of the risk. Character.AI's help center states that "Character.AI does not collect or store any personal data submitted during age verification," that "any biometric data or ID submitted is automatically and permanently deleted within 7 days," and that "Character.AI does not have access to your submissions at any point." It also says you do not have to ask: "deletion is automatic. You do not need to contact us to have your data removed." Persona, it notes, is "used by companies like LinkedIn, OpenAI, Block, and Etsy."

Take that at face value and the exposure is a seven day window at one specialist vendor, not a permanent record at a chat company. That is a materially better deal than most people assume when they see a face scan prompt, and it is more than almost any competitor in this category has put in writing. It is worth weighing honestly rather than dismissing.

What it does not do is remove the underlying asymmetry. A password can be rotated after a breach and a face cannot, so even a short retention window is a different class of risk from an email address, and the promise is a policy rather than something you can audit from outside. Handing over a document to keep access to a roleplay chat is a defensible choice. It should just be a deliberate one rather than something you click through because the flow put it in front of you.

If you decide to go through with it, do the boring things: verify you are on the real character.ai domain rather than a lookalike prompted by an email, complete the check in the app rather than from a link someone sent you, and treat any request for an ID that arrives by message as fake. Verification prompts appear inside the product, not in your inbox. And if a document upload is what finally makes you look at your wider exposure, the more useful move is to reduce the personal data already sitting with brokers, because that is where a leaked record actually gets combined with everything else about you. You can get your personal information taken down from data broker sites in a way you cannot undo an ID upload.

Is Character AI safe to use?

For an adult, in the ordinary sense, yes. It is a well funded company with a named legal entity, a real safety program, an independent safety lab it pays for, and more regulatory attention on it than any other product in this category. Nothing about using it puts your device or your money at risk. It is not a scam and it never was.

Safe is doing a lot of work in that sentence, though. There are three different questions hiding inside it. Is the company legitimate: yes. Is your data private: no, and specifically not in the US. Is the experience safe for a vulnerable person: that is the question the litigation was about, and the company's own answer was to remove the product from minors entirely rather than to argue it was fine.

Is Character AI safe and private?

Legitimate, not private. Character.AI's own model training page states that user generated content "can help us improve the creative writing qualities and functionality of our models." The opt out exists, and it is offered to users in the European Economic Area and the United Kingdom. If you are in the United States, that toggle is not for you. We compared how every major app handles this, including who owns what your companion writes, in our guide to the best AI companion app for privacy.

Read the rest of that page too, because it is unusually candid. Even after opting out, Character.AI says it "still may use your data to improve other aspects of our services like search, recommendations, and safety classifiers." So the European opt out narrows what your chats train. It does not remove them from the pipeline.

Set that against the category baseline. The Mozilla Foundation's February 2024 Privacy Not Included review looked at 11 romantic AI chatbots and gave a warning label to all 11, with 10 of the 11 failing its minimum security standards. Character.AI was not in that review, so it was neither cleared nor condemned. What we can say is that its published position is more transparent than most and its actual practice, no US opt out, is standard for the category rather than better than it.

Does Character AI save all your chats?

Yes. Conversations are stored on company servers, they persist across sessions, and automated systems scan them for policy violations with human escalation when something is flagged. That is not unique to Character.AI and it is not sinister, but people consistently underestimate it. The working assumption for anything you type into any companion app should be that a human could read it if a classifier decides they should.

Deleting a conversation removes it from your view. It does not reliably remove it from backups, from moderation logs, or from a training set it has already contributed to. This is the practical reason not to put material into a chatbot that you would not want associated with your name: not because anyone is reading it today, but because deletion is a weaker promise than it sounds.

Is Character AI safe for adults?

It is safe for adults and increasingly awkward for them. The whole platform has been reshaped around the assumption that a given user might be a minor, which is the right call for the company and a worse product for everyone over 18. Filters that were already the most common complaint got tighter. An adult can be misclassified by the age model and locked out of an account with years of history in it until they hand over a document.

There is also the ordinary question of what you are buying. Character.AI is a character roleplay platform, not a companion built to hold a relationship over time. It runs on a context window rather than a durable memory store, which is why long running conversations drift and characters forget things you told them last week. If that is the actual problem you are trying to solve, no amount of safety improvement fixes it, and it is worth understanding how AI companion memory actually works before you pay anyone. We laid out the full picture in our Character AI review, including what c.ai+ does and does not unlock.

What happens if you fail Character AI age verification?

The account is restricted rather than deleted. Open ended chat becomes unavailable, adult oriented characters disappear, and what remains is the creation tools and read only access to conversations you already had. If you are an adult who was misclassified, the route back is the Persona check, and passing it restores the account.

If you decline to verify, you keep a reduced account indefinitely. That is worth knowing before you start: there is no timer forcing a decision, and nobody deletes your characters because you said no. Plenty of adults have simply stopped there and moved on rather than upload a document, which is a reasonable place to land.

What is Character AI reading mode?

Reading mode is the restricted account state. Character.AI defines it as "the experience for users who are verified as under 18 on Character.AI," where "you can browse characters and view content on the platform, but chat features are not available." It is what you land in if the check places you under 18, and since the 14 July 2026 help page update it applies to accounts verified as under 18 rather than to everyone awaiting a check.

More survives than people expect. The help center lists four things reading mode keeps: "all past chats remain saved and viewable," creating characters, scenes and videos, the Imagine Chat and AvatarFX features, and feed activity including "sending stickers, reacting, and supporting creators." So the creation side of the product stays open and the conversation side closes.

For an adult who was misclassified, that is the practical cost of saying no: you keep the archive and the tools, and you lose the back and forth. Passing the Persona check is the only route out of it, and it restores the account rather than rebuilding it.

Is Character AI safe from hackers?

There is no publicly documented breach of Character.AI user data, which is the most anyone can honestly say. Absence of a reported breach is not proof of good security, and no independent researcher has published a security audit of the platform. What we know is that it is a large, well capitalized company with the security posture that usually implies, and that it now routes the most sensitive part of the flow, identity documents, to a specialist vendor rather than storing them itself. That last decision genuinely reduces your exposure.

How does Character AI age verification compare to other AI apps?

It is the strictest of the major players, and that is the honest answer even though it is the one people least want to hear. Character.AI scores your account with an in house model and escalates to Persona for an ID or a face scan. Most of its competitors ask for nothing at all, and one of the largest asks for a birth year you can scroll.

The clearest outside benchmark published in 2026 comes from the Common Sense Media Youth AI Safety Institute, which assessed Grok on 22 January 2026 and rated it unacceptable risk. It found xAI's terms set a minimum age of 13 with parental permission required for 13 to 17 year olds, that the Grok website performs no age verification at all, and that the app relies on self reported age that is easy to circumvent. Its summary of the gap is worth quoting because it is the exact failure Character.AI's check is designed to close: teens who provide an adult birth date get the full adult experience with no restrictions.

A second 2026 benchmark points the other way, and it is the more useful one for an adult deciding where to spend time. Janitor AI began asking for age verification on 24 April 2026 in Australia and Brazil, and on 15 June 2026 in the UK, but only for users connecting from those countries. Its help center is explicit that anyone outside those regions is completely unaffected, and the company said it has no plans to roll the check out anywhere else. A US adult on Janitor is asked for nothing.

That contrast is the whole decision in one line. Character.AI applies age assurance to everyone, everywhere, on its own initiative. Janitor applies it only where a government forced it, under threat of fines up to $35 million in Australia, $9.5 million in Brazil or £18 million in the UK, and said plainly that it is fundamentally against KYC for social media. Neither position is unreasonable. One optimizes for closing the teen access gap, the other for not collecting documents it was never asked to collect. We went through Janitor's side of it, including what its k-ID flow actually retains, in our review of whether Janitor AI is safe.

Set against that, an ID check looks less like an overreach and more like the only mechanism anyone has that actually works. The trade you are being asked to make is real either way: Character.AI takes a document or a face scan and closes the gap, while the apps that ask for nothing leave it wide open and hand you the convenience instead. There is no option on the table that gives you both.

There is a further wrinkle that almost nobody checks, and it cuts against Character.AI. Every one of these apps has to declare its own content to Apple, and those declarations are public. Checked on 1 September 2026, Character.AI's US App Store listing discloses only Infrequent/Mild Sexual Content and Nudity across 555,372 ratings. PolyBuzz, from Cloud Whale Interactive Technology LLC, discloses Frequent/Intense Sexual Content or Nudity across 468,619 ratings, and asks for no ID at all. Both carry the same 17+ rating.

Read that plainly and the burden in this category is close to inverted. The app that verifies hardest is disclosing less mature content than a similarly sized rival that verifies nothing, so the check you are being asked to pass is not proportional to what is actually behind it. That is not an argument against age assurance, and Character.AI's position is defensible on its own terms. It is an argument that a face scan at the door tells you very little about what the room contains, and that comparing apps on their verification flow alone will mislead you. We built the cross category version of that comparison, using each developer's own disclosures, in our guide to the best AI companion apps for adults.

Worth knowing if you are comparing on price as well as policy: the companion features on Grok sat behind SuperGrok at $30 a month, and xAI said on 24 July 2026 that it is retiring the avatar companions entirely. We went through what that means for anyone currently paying for it on our page on Grok Ani alternatives. For the wider cross brand picture of who demands what, we compared the checks row by row in our guide to the best AI companion app without ID verification.

Is Character AI safe anymore?

It is safer than it was, and that is precisely why it suits adults less well. Every change since October 2025 has been aimed at the failure mode that generated the lawsuits, and those changes worked in the direction they were pointed. The cost was paid by adult users in tighter filters, age checks and an experience built around the possibility that you are 15.

So decide on what you actually want. If you want character roleplay and you do not mind the filter, stay; it remains the best library of characters anywhere and c.ai+ is cheap. If the problem is that the filter keeps interrupting the thing you came for, or that your companion forgets you, or that you would rather not train someone's model with your private conversations and have no opt out because you live in the wrong country, then the platform is not going to change back and you are better off looking at an alternative to Character AI built for adults, or at the wider field of apps like Character AI ranked by what each one demands from you. The same three questions are worth asking of whatever you move to, and we ran them on the other big names in Character AI compared with Replika and in our look at whether Replika is safe.

Amorous.ai charges $9 a month billed annually or $12 month to month, flat, with nothing metered. We are an adults only product, so we verify age at signup rather than retrofitting a check onto accounts, and we do not train models on your conversations. We also do not have Character.AI's character library and we are not trying to; it is a different product for a different job, and if character roleplay is what you want, it is the better purchase.

Every fact, date, price and quotation on this page comes from a primary source: Character.AI's own company blog, help center and model training page, the help center's published article vote counts, court reporting from CNBC, Axios, the Washington Post, CBS News and NPR, and the text of California SB 243. Where the help center is quoted, the wording is verbatim.