Google Gemini was not built for children, and whether it is safe for yours depends on which Gemini they have. The version inside a school Google account has stronger data rules than the one at home, but Common Sense Media rates both "High Risk" for under-18s, and no filter can tell when a child is struggling. Your school's settings and your own rules decide more than Google does.
Last updated October 19, 2026. I re-read Google's Gemini help page, its Family Link page, its Classroom admin page and the student hub announcement today. What changed: Google's Gemini help page now says a Family Link account still cannot use the Gemini website, the exact Family Link steps are below, the August change has its announcement date, and I corrected one line about what Common Sense Media recommends for teens. The platform facts here move every few weeks, and each one carries the date I checked it.
The short answer, by age
| Age | Verdict | The one thing to do tonight |
|---|---|---|
| Under 8 | Not alone, at school or at home | Ask the school whether the Gemini tab is on for your child's class |
| 8 to 12 | Only with you beside them at first | Ask it something you already know the answer to, together |
| 13 to 15 | Homework and creative projects yes, feelings no | Agree the purpose rule out loud |
| 16 to 18 | Fine as a tool, unreliable as a judge | Teach them that it backs down when pushed |
This is my reading of the evidence below, not a rating from Google or Common Sense Media. A school's rules, Google's age limits and a doctor's or safeguarding lead's advice all come before it.
Which Gemini does your child actually have?
Most children now have access to two different Geminis, and they play by different rules. Before you can answer the safety question, you need to know which one you are asking about.
The school one. Since June 20, 2025, Google has made the Gemini app available to Google Workspace for Education users "of all ages", with what it describes as stricter content policies for younger users, a youth onboarding experience and a double-check feature that runs the first time a student asks a factual question. In an announcement dated August 4, 2026, Google said it would go further: a Gemini tab inside Google Classroom for students of every age, from the youngest primary (elementary) classes upwards, rolling out on the web from August 10 and on mobile from August 17, on by default wherever the school had already allowed the Gemini app or NotebookLM. The same announcement says that where a school had turned Gemini in Classroom off for students, "access will remain off for all students, including students younger than 18". Google's own admin help page puts the default plainly: "Gemini in Classroom is ON by default for users of all ages." Checked October 19, 2026, it gives the administrator's path as Generative AI, then Gemini for Workspace, then Feature access in Classroom, with On or Off for each group. The school has to make that choice. If your child logs into Google Classroom, the fair assumption in October 2026 is that Gemini is there unless the school has said otherwise.
The home one. Google's Gemini help page, which I read again on October 19, 2026, says: "You must be 13 (or the applicable age in your country) or over to use Gemini with a personal or school Google Account." That means 13 in the United States and the United Kingdom and higher in several European countries. The same page says: "You still can't access the Gemini web app with a Google Account managed by Family Link." Google's Family Link help page, also checked October 19, 2026, lets a parent switch on Gemini Apps for a child under 13 (or the applicable age in your country), says the parent receives an email notification when the child first opens it, and adds that access is being released gradually, so a child may not have it yet even after you switch it on. Read together, my reading is that a supervised child's home access is through the Gemini app on a phone rather than the website. Google does not say that in one sentence, so treat it as my interpretation. And the Family Link page still states that "Gemini Apps are not available for supervised accounts in the European Economic Area, Switzerland, and the United Kingdom", so a British under-13 does not have a home Gemini account at all. A British teenager does.
To check or change it yourself on a supervised account, open the Family Link app (or go to familylink.google.com), select your child, tap Controls, then Gemini, then Gemini Apps, and switch it on or off. In older versions of the app the path is Controls, then Content restrictions, then Gemini, then Gemini Apps. That is Google's own wording of the steps, checked October 19, 2026, and menus move, so if yours looks different, search Family Link's help for "Gemini".
One wrinkle I cannot resolve for you. Google's Gemini help page says 13 or over for a school account, while its Workspace for Education announcements say students of all ages. Google does not reconcile the two on one page. My reading is that the Workspace announcements and admin pages govern an account a school manages, because they are written for those accounts, but the practical answer is to ask the school what is switched on and for which year groups rather than rely on either page.
The school version has one clear advantage. Google's education privacy page states that chats on a Workspace for Education account "are not reviewed by human reviewers or otherwise used to improve generative AI models". Common Sense Media's testers agreed that the data side is done properly. The problem, as the next section shows, is everything the data rules do not cover.
What Common Sense Media found, and what "High Risk" means
Common Sense Media has tested Gemini three times in a year, and the rating has not moved. Its Gemini rating page, which I checked on October 19, 2026, still shows an overall risk level of High, last updated September 5, 2025. In September 2025 its assessment of Gemini's under-13 and teen versions rated both "High Risk", finding that they were essentially adult Gemini with added filters, that they treated every child the same regardless of age, and that they could still surface content about sex, drugs and alcohol and give unsafe mental health guidance. The page's own words are that Gemini "far too easily provides unsafe mental health and emotional support". In January 2026 its Youth AI Safety Institute rated the school version, Gemini K-12, "High Risk" as well. That report praised the data protection and noted that Gemini says plainly that it is an AI and refuses role-play, but found that it did not always recognize or respond appropriately to signs of a child in difficulty, that it "backs down when challenged", and that its answers average an 11th-grade (Year 12) reading level, too hard for most of the children it is now switched on for. In August 2026, after the Classroom change, the same institute reported that the Guided Learning mode "is easy to switch out of" and will write full essays when asked.
Here is what I make of that, as my own reading rather than theirs. "High Risk" does not mean a child who opens Gemini is in danger. It means the product has not been designed around a child, and the failures cluster around the two things no filter can do: know how old the person typing really is, and notice when they are not all right. Everything a family does at home is about covering those two gaps.
What just changed at school, and why nobody told you
The change reached schools before it reached parents, and in many cases before it reached the teachers. Education Week reported on September 1, 2026 that at least one principal had learned about the Classroom rollout from news coverage rather than from Google or his district, and that his Ohio high school switched it off because the staff had not been trained on it. Chicago Public Schools, Axios reported on August 24, 2026, had dropped its planned Gemini rollout to high school students after its own AI committees raised concerns about readiness. New York City went further for its younger pupils: on September 2, 2026 the Mayor's Office announced a one-year pause on student-facing generative AI for the 2026-27 school year, covering the early grades through 8th grade, and what that pause does and does not cover at home is worth reading beside this. Common Sense Media's Robbie Torney described the Gemini rollout as defaulting "to adoption rather than permission", which is the fairest one-line summary I have seen.
Parents were already in the dark before this. In the National Parents Union survey of 1,511 US public school parents, fielded by Echelon Insights from February 12 to 18, 2026, 47% said their child's school had not provided information about its AI policy that school year; 37% said it had. In the same survey, 79% supported requiring a parent's permission before a minor uses an AI chatbot. And in RAND's nationally representative survey of American students in spring 2025, more than 80% said their teachers had not explicitly taught them how to use AI for schoolwork. The UK picture rhymes: Internet Matters found in July 2025 that two thirds of UK children aged 9 to 17 had used an AI chatbot, 42% of those users for schoolwork, and only 57% had discussed AI with a teacher.
A default setting is a decision somebody else made for your child. Your job is not to reverse it. Your job is to know it was made.
What Google added in September and October
The August change was not the end of it, and this is the part no school newsletter has caught up with yet. Two more Gemini features reached school accounts in the weeks after, both of them for pupils of every age.
On September 22, 2026 Google announced study notebooks in Gemini, which turn a set of materials into lessons, quizzes and a running score of what a child has mastered. Google's wording is that "users of all ages who are signed into a school or work-issued Google account will also have access to study notebooks". This one is not on by default. A pupil gets it only where the administrator has both Gemini Notebook and Gemini set to On for their group, and the gradual rollout started on September 17, 2026.
Then on October 1, 2026 Google opened a student hub at gemini.google.com/students, gathering the study notebooks, flashcards and practice quizzes into one place and making it "available to Google Workspace for Education users of all ages". On this one Google says plainly that "there is no end user setting for this feature". Whether your child has it is settled by the Generative AI settings in the Workspace admin console, and nowhere else. I re-read that announcement on October 19, 2026 and found nothing newer from Google on either feature.
One geographic detail is worth knowing, because it runs the opposite way to the home rule above. Google's availability line for both features reads "outside of the European Economic Area (EEA)". The United Kingdom is not in the EEA. Read together with the Family Link page, that means a British child who cannot have Gemini at all on a supervised home account can have the full set of Gemini study tools on their school account, while a child in Dublin or Berlin is still waiting for them. That is my reading of the two documents rather than something Google says in one place, and if it sounds backwards it is because the two rules come from different rooms: the home restriction is Google's own caution about supervised accounts, and the school rollout follows the contract the school signed.
What I make of the pattern, as my view rather than theirs: the pace is the story. Three changes in about eight weeks, each arriving through an admin console rather than a letter home, is faster than any school communicates. You are not going to keep up with it, and you should stop trying. Ask the questions in the next section once a term instead, because they work whatever Google shipped that month.
What to Ask the School
I have sat on both sides of this. As an assistant headteacher, and now as a trustee at Stanwix School, I have been in the meetings where a school decides what to switch on for children, and the safeguarding meetings that follow when something was switched on without one. The pattern I see most often in the schools I work with is that a tool like Gemini arrives through the admin console, not through the curriculum: one person accepts a default, teachers find out when a pupil shows them, and the safeguarding lead finds out last. Nobody is at fault. The rollout was designed to work that way. Four questions cut through it, and a good school will not mind being asked.
1. Is Gemini switched on for my child's account, and who decided? Why it matters: the answer tells you whether anyone made a decision at all. A good answer names a person and a reason ("we enabled it for Years 10 and above after the department heads reviewed it"). "It came with Google" is an honest answer, and a sign the school has work to do.
2. Have the teachers been trained on it? Why it matters: Common Sense Media found that Gemini's lesson-planning tools can produce biased material without careful review, and its student mode can be talked into doing the work. A good answer describes what training happened and when. This is the question I would ask first at parents' evening. If you also want to know whether AI is touching the marking rather than the learning, that is a separate question with a separate answer.
3. What happens if my child types something worrying into it? Why it matters: I can find no parent alert anywhere in Google's documentation for the school version, and Common Sense Media notes that administrators cannot stop students raising personal or emotional topics with it. Some school monitoring systems read web searches but not what is typed into a chatbot. A good answer tells you whether the school's monitoring covers Gemini and who would see a flag. Under KCSIE in England that sits with the designated safeguarding lead; in the US, ask the counselor or the technology director.
4. Can I opt my child out? Why it matters: Google's control works by group, not by family, so a school can switch Gemini off for a year group or a class but not easily for one child. The honest answer may be no. What you can ask for instead is that the school tells parents what is on, for whom, and why, which is the standard the National Parents Union figures show most schools are not yet meeting.
What I would do at home tonight, by age
Wherever the school has landed, the home rules are yours. The framework I use is the same one as for ChatGPT's parental controls: controls work in three layers: the account, which is what the app allows; the device, which is what the phone allows; and the conversation, which is what your child tells you. Only the third layer covers the logged-out browser. With Gemini the account layer is split between you and the school, which is why the conversation layer matters more here than for any other tool.
Under 8. My view is that a child this age should not be using Gemini alone, at school or at home, and that close supervision is the right default for years afterward. If your child's primary school has the Classroom tab on for this age, ask question one above, and ask it kindly but clearly. At home there is nothing to set up in the UK, because the supervised account cannot have Gemini; in the US, leave the Family Link Gemini Apps toggle off.
8 to 12. This is the age most affected by the Classroom change, because it is the first time many of these children have had an open chatbot in their school account. Sit with them the first few times. You could say: "Show me what the Gemini tab does on your Classroom. Let's ask it something we already know the answer to and see if it gets it right." That one exercise teaches the three checking questions better than any lecture: Who says? How could we check? What would it look like if this were wrong?
13 to 15. Now there are two Geminis, and the rule that works is about purpose rather than time. Common Sense Media's Gemini rating says "No AI for companionship (including mental health or emotional support) for any teen under 18", and notes that the tool may be helpful for homework or creative projects as long as the teen does not use it for emotional support. That is the line I would draw at home. You could say: "Gemini is fine for homework and for making revision cards. It is not the place to take a bad day. If you ever find yourself telling it things you would not tell me, that is the moment to come and find me, and I promise not to make it weird." For homework itself, the Homework Test still applies: if the machine did this part, what would my child lose? And if the pull is company rather than homework, the Substitute Test is the question I ask: is it adding to my child's life, or standing in for something that is missing?
16 to 18. By this age the safety conversation is about quality, not access. Tell them the specific thing the testers found: Gemini backs down when challenged, which means a confident teenager can talk it out of a correct answer as easily as into a wrong one. You could say: "When it agrees with you the second time, that is not evidence. Ask it what the strongest case against you is." That habit will serve them at university and in work long after the tool has changed.
If your teenager is using any chatbot for their feelings, that is a conversation to have, not a crisis in itself. If you are worried about their safety, the GP or pediatrician, the school's safeguarding lead, 988 in the US or Samaritans on 116 123 in the UK come before any setting on any app. A chatbot is not a clinician, and Common Sense Media's testing is a reminder that it may not notice when one is needed.
What no setting can guarantee
Turning Gemini off at school does not remove it from a personal account, a friend's phone or a logged-out browser. Turning it off at home does not remove it from the school account. Google's under-18 filters reduce inappropriate content; they do not read a child's mood, and by Common Sense Media's testing they do not reliably recognize a child who is struggling. Google's double-check feature runs automatically only the first time a factual question is asked; after that it is a button your child has to choose to press. Your school's monitoring may or may not see chatbot conversations, and you will only know by asking.
And one piece of Google's AI is not covered by any setting at all. In July 2026 Common Sense Media assessed the AI answers built into Google Search, AI Overviews and AI Mode, and rated them an "unacceptable" risk for children. The sentence that matters for a family is its summary of the practical problem: "Unlike Google's standalone Gemini chatbot, schools and parents can't turn off Google Search's AI answers." The same assessment found that "AI Mode completes kids' homework assignments for them every time", and that "because AI Mode is built into school-issued laptops, students can route any assignment through an AI shortcut, and schools have no setting to stop them". So a school that has switched Gemini off has not switched Google's AI off, and neither have you. That one is a conversation, not a control.
The school's rules win on the school account, because it is the school's account. Google's age limits win on a personal account, and lying about a date of birth to get around them is the one thing I would ask families not to do, because every protection Google adds for teenagers hangs on that date. The same is true on the other tools your child uses, and the stated age limits are not the same on any two of them. A doctor's or a safeguarding lead's advice wins over anything in this post.
Where this leaves you
Gemini is safe enough for schoolwork with a teenager who has been shown how it fails, and not safe enough to be the friend of a child of any age. The change at school this term did not make it more dangerous; it made it more present, and it moved a decision from the classroom to a settings page. Find out which Gemini your child has, ask the school the four questions, and hold to the one principle that outlasts every setting: let AI help with the doing; keep the thinking with your child.
If your child's school switched Gemini on and has not yet spoken to parents about it, this is the work I do with schools: an honest look at how ready a school is for AI, including what its teachers have been trained on and what its parents have been told. It is worth sending to your head of year.
Google will change these settings again, probably before Christmas. My Sunday newsletter keeps more than 44,000 teachers and parents up to date with what moved each week, without the panic. Parents are very welcome.
Dan Fitzpatrick is a former assistant headteacher and school trustee, a Forbes contributor, and the founder of The AI Educator. Read Dan's story.
