2026-08-30 · Sienna Richardson

Is ChatGPT Safe for Kids? An Honest Answer

OpenAI turned on ChatGPT for Teens globally on 18 August 2026, which means the answer to this question changed less than two weeks ago. If you last looked into this in 2025, most of what you read is now out of date - and so is most of what is currently ranking for it.

The short version: ChatGPT is safer for a 15-year-old than it was a month ago, it is not designed for a 9-year-old at all, and the parental controls do one specific thing that parents consistently expect them to do and they do not.

What changed in August 2026

Accounts that OpenAI identifies as belonging to someone under 18 now get ChatGPT for Teens applied automatically, on free and paid personal plans alike. It is not something a parent switches on; it is the default for those accounts.

Teen accounts carry additional content protections. OpenAI describes these as reducing graphic content, viral challenges, sexual, romantic or violent roleplay, and extreme beauty ideals.

Underneath that sits an age prediction system, which tries to work out whether a user is under 18 rather than trusting the birthday they typed. When it is unsure, it applies the teen settings anyway. That default matters: it means a teenager who lied about their age at signup is not automatically outside the protections.

A separate update in July 2026 expanded the safety notifications a parent receives, including alerting a parent when a linked teen's account is banned for violent activity.

The thing parental controls do not do

This is the single most important sentence in this article, and it is the one parents are most often surprised by.

ChatGPT's parental controls do not let you read your child's conversations. You can link accounts, set restrictions, and receive safety notifications in specific situations. You cannot open the app and see what they have been talking about.

OpenAI has been explicit about this, and it is a deliberate design decision rather than a missing feature. If your reason for looking into this was "I want to see what my teenager is asking it", no setting inside ChatGPT provides that, and no version of it is coming as a toggle.

That leaves two honest options. Either you accept conversation privacy as the trade for the safety features, which for a 16-year-old is a defensible position, or you decide the platform is not appropriate for your child yet.

Age by age

The right answer genuinely differs by age more than by anything else.

Under 13. ChatGPT's terms require users to be at least 13, so an account for a younger child is against the rules to begin with. Beyond the terms, the failure mode is not shocking content - it is that a nine-year-old cannot reliably tell a confident wrong answer from a right one, and ChatGPT is extremely confident when it is wrong. For homework help at this age it teaches the wrong lesson.

13 to 15. This is the range where teen protections do real work and where supervision still makes sense. The realistic risks here are academic - work handed in that is not theirs - and dependency, rather than explicit content. Link the accounts, turn on notifications, and be specific about what counts as using it to learn versus using it to avoid learning.

16 to 17. Teen protections still apply, but the case for treating this as a supervised activity gets weaker. A 17-year-old will use AI tools throughout university and work, and the useful parental contribution at this point is judgement, not blocking - how to check what it tells you, when it is confidently wrong, what not to type into it.

The risks that are actually worth worrying about

Ranked by how often they turn out to matter, rather than by how alarming they sound.

It is confidently wrong. Far and away the most common real problem. ChatGPT states incorrect facts in exactly the same tone as correct ones, and children have almost no defence against that because confident delivery is how they have learned to identify a reliable adult. A child using it for homework absorbs errors with no signal that anything went wrong.

It replaces the thinking, not the effort. A child who asks for an explanation is learning. A child who asks for the answer is not, and the two are indistinguishable from the outside until a test happens. This is the one worth having an explicit rule about.

Personal information. Children share things with a chatbot they would not put in a group chat, because it does not feel like an audience. Names, schools, addresses, photos of homework with a name in the corner. The rule that works is simple and absolute: nothing you would not put on a poster.

Dependency. Less acute for ChatGPT than for companion apps built to be a friend, but not zero. A chatbot that is always available and never impatient is an easier place to take a worry than a person is. Worth noticing if it becomes where your child goes first.

Explicit content. Genuinely the least likely of these on ChatGPT specifically, and the one most articles lead with. The teen protections target it directly and it is not what the product is for. If explicit AI content is your concern, the platforms it applies to are companion and roleplay apps, not this one.

ChatGPT compared to the other AI apps kids use

Platform	Minimum age	Teen protections	Parent can read chats	Main risk
ChatGPT	13	Yes, automatic since Aug 2026	No	Wrong answers, dependency on it for schoolwork
Claude	18	Not built for minors	No	Not intended for children at all
Character.AI	13	Open-ended chat banned for under-18s since Nov 2025	No	Was emotional dependency; the ban changed this
Google Gemini	13, with Family Link	Yes, via Family Link	No	Similar to ChatGPT
Smaller companion apps	Varies, often unenforced	Frequently none	No	Explicit content, no meaningful filtering

The pattern worth taking from that table: the large, well-known platforms have invested heavily in teen safety and are broadly the safer choice. The risk concentrates in the small apps nobody has heard of, which have no age enforcement, no content filtering worth the name, and no parental controls at all. More on the AI chatbot landscape here.

What to actually do

Link the accounts. Parental controls in ChatGPT need both sides connected. Do it together rather than covertly - covert monitoring discovered later costs more than it gains, particularly with a teenager.

Check the birthdate on the account. Every protection above keys off the age. A child who signed up claiming to be 20 may be outside the teen defaults, although the age prediction system now catches a lot of these.

Make one rule about homework and make it specific. "Ask it to explain, don't ask it to write" is a rule a 13-year-old can actually follow. "Use it responsibly" is not.

Decide about the small apps separately. ChatGPT and Character.AI are the ones with safety teams. The decision about Chai, Janitor AI and the rest is a different decision, and for most families a much easier one.

Where 3Eyes Fits

Allowing ChatGPT while blocking the apps with no safety review at all is a reasonable position, and it is awkward to implement with a blocklist - new AI apps appear weekly and a blocklist has to have heard of one before it can stop it.

3Eyes handles it the other way round. Every site is blocked until you approve it, so you approve chatgpt.com for homework and every companion app that launches next month is already blocked, without you having to know its name. When your child wants something new they ask from the blocked page and you approve or decline from your phone in seconds.

It also does the part that matters for the homework problem specifically: the machine carries educational games, learning activities and a built-in tutor, so there is something on it that teaches rather than answers. Setup is about two minutes. See how it works.

Frequently asked questions

What age is ChatGPT for? OpenAI's terms set the minimum at 13, and under-18s are expected to have parental permission. Accounts identified as under 18 get teen protections applied automatically.

Can I see my child's ChatGPT conversations? No. Parental controls provide restrictions and safety notifications, not transcripts. This is deliberate, and there is no setting that changes it.

Is ChatGPT safe for a 10-year-old? It is below the minimum age, and the practical problem at that age is that a child cannot judge whether an answer is right. If you want AI help for a younger child, a tool built to teach a curriculum is a better fit than a general chatbot.

Will I be told if something goes wrong? In specific situations, yes. OpenAI expanded parent safety notifications in July 2026 to cover more circumstances where a teen may need urgent adult support, including a linked account being banned for violent activity. It is not a general activity report.

Should I block ChatGPT? For most teenagers, no. It is going to be part of their education and their work, and a blanket block mostly relocates the usage to a friend's phone. Rules about homework plus linked accounts get you further. For under-13s, blocking is reasonable, and so is choosing a tool built for children instead.