My sister-in-law called on a Tuesday night, kids in the background, voice low like she didn’t want them to hear. Her twelve-year-old had been using something called Character.AI and she’d only just found out. She wanted to know: what’s the minimum age? Can she see what he’s been typing? Is there a way to just — turn it off? I started to answer and then stopped myself, because the answer I’d given my neighbor two months earlier was already wrong. These platforms had changed their policies again. I told her I’d send her a link, then spent forty-five minutes realizing the link I wanted didn’t exist yet.
So I made it. What follows is a single-page reference — one H2 per platform, each section self-contained — compiled and last verified in July 2026. I’ll update it when things change, and they will change; this landscape shifts roughly monthly. Bookmark it, send it to your sister-in-law. Everything here comes from official policy pages and support documentation, linked directly. I am not a lawyer or a child development researcher. I’m a parent who reads the policy updates so you don’t have to do it at midnight.
Before we get into platforms: a Pew Research Center survey fielded in fall 2025 and published February 24, 2026 found that a majority of US teens now use AI chatbots, with about 30% using them daily. That number reframes everything. We’re not talking about a niche activity. We’re talking about something as ordinary, for a lot of teenagers, as texting.
OpenAI / ChatGPT
Minimum age: 13.
Parental controls, as of July 2026: OpenAI launched parental controls on September 29, 2025. A parent links to the teen’s account via email invite and can set quiet hours, disable voice mode, memory, image generation, and model training. Teen accounts automatically get reduced graphic content and limited romantic or violent roleplay. If the system detects signs of acute distress or self-harm risk, parents are notified.
What changed recently: Around January 20, 2026, OpenAI rolled out age prediction — behavioral signals estimate a user’s age, uncertain cases default to the under-18 experience, and adults can verify to unlock full features. This means the system is actively trying to catch teenagers who registered with a false birthdate, not just relying on self-reported age.
What parents can’t do: Read conversation transcripts. The parental dashboard surfaces controls, not content.
Character.AI
Minimum age: 13 to register.
Parental controls, as of July 2026: Since November 25, 2025 — announced October 29, 2025 — users under 18 no longer have access to open-ended chat at all. The under-18 experience is restricted to creative-only features: guided “Stories,” character and video creation, and a community feed. Age assurance runs on all accounts. Users flagged as minors who believe they’re adults can verify through a third-party service called Persona.
What changed recently: This is the most dramatic policy shift of any platform on this list. Where Character.AI was once essentially an unmoderated character-roleplay chatbot accessible to anyone, it now offers teenagers something closer to a structured creative tool. The open-ended chat that generated so much parental alarm — and legal scrutiny — is simply gone for under-18 accounts.
What parents can’t do: There is no parent-linking feature analogous to ChatGPT’s. The protection here is architectural — the risky feature is removed — rather than supervisory.
Google Gemini
Minimum age: Under 13 can now access Gemini — but only through a Family Link-supervised Google account, a policy that went into effect in May 2025. At 13, the child can opt to self-manage their account.
Parental controls, as of July 2026: Through Family Link, a parent can toggle Gemini access off entirely (Family Link > Controls > Gemini). The parent receives an email the first time the child uses it. Voice Match is not available for supervised accounts under 13. Gemini is not offered to Family Link-supervised accounts in the EEA, UK, or Switzerland.
What changed recently: The extension to under-13 users is the notable development here. Google has moved toward allowing younger children in, with the tradeoff being that Family Link supervision is required and the parent has a clear off-switch. Whether that tradeoff is acceptable is a judgment call every family has to make.
Meta AI (Instagram and Facebook)
Minimum age: Tied to the platform minimum — 13 for Instagram and Facebook in the US.
Parental controls, as of July 2026: Meta announced its approach on October 17, 2025, with rollout on Instagram beginning in early 2026 in the US, UK, Canada, and Australia. Parents supervising Teen Accounts can turn off one-on-one chats with AI characters entirely or block specific characters. The main Meta AI assistant remains on, with age-appropriate, PG-13-guided defaults. Meta AI will not discuss suicide, self-harm, or sex with teens and redirects to resources instead.
What changed recently: Since April 23, 2026, parents in the US, UK, Australia, Canada, and Brazil can see topic categories — not full transcripts — covering what their teen discussed with Meta AI over the past seven days. This is a meaningful middle ground: parents get a topical map without reading every message.
Anthropic / Claude
Minimum age: 18, full stop. Claude is the only major general-purpose AI assistant on this list with an 18+ minimum.
Parental controls, as of July 2026: None exist, because the platform does not permit anyone under 18. There is no kids’ mode, no family linking, no supervised experience. Per Anthropic’s support documentation, enforcement tightened in 2026: accounts of confirmed minors are disabled, and age verification can be required to restore access.
What this means practically: If your teenager is using Claude, they either lied about their age during signup or are accessing it through an employer or school account. That’s worth knowing — not because it makes Claude uniquely dangerous, but because it means any safeguards are entirely on your end, not the platform’s.
Can I ask you something? When did you last actually sit down and talk with your kid — not about rules, not about screen time limits, but about what they’re actually doing in these chats, and why?
What we actually do in our house
Our daughter is two, which makes me the wrong person to tell you how to handle a teenager and AI. I know that. What I can tell you is why I wrote this post anyway: because the relatives asking me these questions at holiday dinners are real people with real kids who are using these tools right now, and they deserve accurate information, not vague anxiety.
In our house, we’ve chosen to keep AI out of our daughter’s life for now — not because I think it will permanently harm her, but because she’s two and she doesn’t need it, and I want to understand the landscape before it’s in front of her. That’s a choice we’re able to make at this age. I’m under no illusion that it scales forever, or that our approach will look the same when she’s twelve. What I’m trying to build right now is the habit of paying attention — reading the policy updates, knowing what changed, being the parent who actually checks.
What’s hard: keeping up. Seriously. I drafted an earlier version of this post in April and had to revise three sections before publishing because the platforms had updated their policies in the intervening weeks. I also find it genuinely difficult to know how much weight to give corporate safety announcements versus independent research. I try to link to primary sources and let you decide.
Questions I keep getting
My kid is 13. Which of these platforms is the most restricted for their age?
As of July 2026, Character.AI has made the most dramatic structural change: users under 18 cannot access open-ended chat at all. The platform is now a guided creative tool for teenagers, not a freeform chatbot. That’s not a parental control you configure — it’s a hard architectural limit. ChatGPT’s parental controls are the most granular in terms of what a parent can actually toggle (quiet hours, voice, memory, image generation), but they require the parent to set up the linked account.
Can I read my kid’s AI chat history?
Not through any of the parental dashboards described above — none of them surface transcripts. Meta AI is the closest: since April 23, 2026, parents supervising Teen Accounts in the US, UK, Australia, Canada, and Brazil can see topic categories for the past seven days, but not the actual messages. If you want to know what your kid is talking about with an AI, the most reliable method is still the oldest one: ask them, and make it a conversation they don’t dread.
Is there any AI assistant my under-13 child is actually allowed to use?
As of July 2026, Google Gemini is the only assistant on this list with a supervised access path for children under 13, through Family Link. The parent can toggle access off, is notified on first use, and the child cannot use Voice Match. Note that this access is not available in the EEA, UK, or Switzerland even with Family Link. Every other platform on this list requires the user to be at least 13, with Claude requiring 18.
What do these controls actually not cover?
They don’t cover a teenager who accesses any of these tools on a friend’s account, a school device with different settings, or a platform you haven’t heard of yet. They don’t substitute for a kid who understands why you’re paying attention — not just that you are. Every expert framework I’ve read on this eventually arrives at the same place: the controls buy you time and reduce casual exposure, but the actual protective factor is the relationship and the ongoing conversation. I believe that. I also believe this reference page is more useful than vague reassurances, which is why I wrote it.
