Mostly no. In Gallup's survey of 2,232 US public school teachers, published in June 2025, 2 percent used AI daily to grade or give feedback on student work and 7 percent weekly. Where it does happen, it is usually helping write the comment rather than deciding the grade. The caveat: most teachers have been given no rule about this either way.
That last part is the bit worth your attention, and it is not the part the headlines cover.
What the numbers actually say
Gallup asked 2,232 teachers in US public K-12 schools how often they used AI for ten different jobs. Fieldwork ran from March 18 to April 11, 2025, and the results were published on June 24, 2025. For grading and giving feedback on assignments or assessments, 2 percent said daily, 7 percent weekly and 7 percent monthly. So roughly one teacher in six touches it even occasionally, and five in six do not.
Among the teachers who do use it there, 79 percent said it saves them time on that task. On quality, 13 percent said the feedback was of much higher quality than what they would have produced alone and 44 percent said somewhat higher. That is teachers rating their own output, not an independent measure, and it is worth reading as what it is.
Then a second Gallup survey, of 2,069 US teachers with fieldwork from February 9 to March 2, 2026 and published on May 26, 2026, asked a different question: whether anyone had told them what they were allowed to do. Just 18 percent of US K-12 teachers receive formal guidance on how AI should be used at work. Forty-eight percent get informal guidance only. And on this specific job, the survey found that "58% say the same for how they should use AI for grading and providing student feedback," meaning they receive no guidance at all.
Both surveys are US teachers. I have not found an equivalent national figure for the UK, so treat the numbers as American and the pattern as probably wider.
Here is my reading of the two together, and it is my reading rather than Gallup's finding. The story is not that teachers are quietly handing your child's work to a chatbot. It is that a small number are using it, a great many are unsure whether they may, and almost nobody has been given a line to hold. A practice that nobody has ruled on tends to grow quietly and unevenly, which is a worse outcome than either a clear yes or a clear no.
The difference that matters: the comment and the call
If you only take one thing from this page, take this distinction, because it is the one that decides whether you should mind.
The Comment and the Call. AI can help a teacher write the comment. Only a teacher can make the call. The comment is the feedback your child reads. The call is the judgment about how good the work is and what it means for what comes next. When you ask the school about AI and marking, ask which of the two the software touched.
A teacher who has read thirty essays, formed a view on each, and then used a tool to turn their notes into full sentences for tired parents at 9pm has done nothing to your child. The judgment was theirs. The words were assisted. That is the overwhelming majority of what I see.
A teacher who pastes an unread essay into a chatbot and copies out the score has outsourced the thing they are paid for. That is rare, and it is a professional problem rather than a technology problem.
Most of what parents worry about lives in the first category and gets talked about as though it lived in the second.
It is worth noticing that this is the same distinction schools apply to children. The line between AI helping with homework and AI doing it is also a question about who did the thinking. It would be odd to hold a fifteen year old to that standard and not a teacher.
What I See in Schools
I have trained more than 150,000 teachers across more than thirty countries, and AI marking comes up in nearly every room. The question teachers ask is almost never "how do I get this to grade for me." It is some version of "am I allowed to use this, and will I get into trouble."
Three patterns show up again and again.
The first is that marking is the job teachers most want help with, and for an honest reason. It is the largest single call on their evenings, and it is the part of the job that gets cut when there is too much of it. A teacher choosing between writing forty thorough comments badly at midnight and writing forty good ones with help at nine is not being lazy.
The second is that the tools have crept in sideways. Most schools I work with did not decide to allow AI marking. A few teachers tried it, found it useful, mentioned it in a department meeting, and it spread. There was no meeting, no policy, and no conversation with parents, because at no point did anyone think a decision was being made. That is how most AI use in schools has arrived so far: not as an announcement, but as a habit that formed before anyone named it.
The third is the one that should reassure you most. In schools where somebody has actually written down what is and is not allowed, the practice becomes visible, modest and checkable almost immediately. The problem is not teachers. The problem is silence.
A school that has never discussed AI marking does not have a policy of not using it. It has a policy of not knowing.
Is AI grading actually worse than a teacher?
Not straightforwardly, and this is where I would push back on the usual parent fear.
In a study published in the journal Education Sciences on July 17, 2026, researchers compared AI and human grading of 139 mathematics assessment responses from Greek primary school students in grades 5 and 6, aged 10 and 11. Agreement between the AI systems and the human markers was good, with an intraclass correlation of 0.777 for individual scores. The AI systems were also more consistent than the teachers, producing less variation across the same range of work.
Two findings in there deserve a parent's attention, and they point in opposite directions.
Consistency is real, and it is not nothing. A human marker at essay forty is not the human marker at essay one. Tools do not get tired.
But consistency is not the same as correct. The same study found that one model, GPT-5.0, showed what the researchers called proportional bias: the gap between its scores and the human scores grew at higher levels of student performance. In plain terms, it agreed least about the best work. If your child is the one producing the unusual, ambitious, slightly odd piece, that is precisely the piece a machine is most likely to misread, and precisely the piece a teacher who knows them would recognize.
This is one study, of 139 math answers, in one country, in one subject, with children of one age. It is not the last word on anything. I include it because it is the closest evidence I can find to the population parents are actually asking about, and because it complicates the story rather than confirming it.
Who is responsible if the mark is wrong?
The teacher. This is settled, and it is worth knowing in plain words before you ever need it.
The Department for Education's guidance on generative AI in education, last updated on August 12, 2025, states that "the quality and content of any final documents remains the responsibility of the professional who produced it and the organisation they belong to, regardless of the tools or resources used." It adds that "any content produced requires critical judgement to check for appropriateness and accuracy."
So "the AI gave that grade" is not an answer a school can give you. Whoever put their name to the mark owns it.
The same guidance carries a second point that almost no parent has been told. It says "it is recommended that personal data is not used in generative AI tools," and that schools "must not allow or cause students' original work to be used to train generative AI models unless they have permission." Your child's essay is their original work. If a school is putting whole pieces of pupil work into a free consumer chatbot, the question is not really about the grade at all. It is about where the work went.
That is a reasonable thing to ask about, calmly, without accusing anybody.
What to ask the school
Four questions, in the order I would ask them. Any teacher or head of year can answer all four, and none of them is a complaint.
1. Does the school have a position on staff using AI to help with marking and feedback? A good answer is a yes or a no with a reason. "We allow it for drafting comments, not for awarding grades" is a good answer. "I'll find out" is also a good answer. Vagueness that never resolves is the one to notice.
2. If it is used, does a teacher read the whole piece and decide the grade? This is the comment and the call, asked directly. You are not asking whether software was involved. You are asking whether a person made the judgment.
3. What happens to my child's work when a tool is used? You are asking whether pupil work goes into a school-contracted system or into somebody's personal chatbot account. Those are very different answers and schools know the difference.
4. What training have staff had on this? Given that 58 percent of US teachers report no guidance at all on using AI for grading, this is usually the real gap, and it is a kinder question than it sounds. It puts the weight on the school's systems rather than on an individual teacher.
You could say, at a parents' evening: "I'm not worried about anyone in particular. I just wanted to know whether the school has a line on AI and marking, so I know what to tell my son when he asks."
How this differs by age
Under 8. This is not a conversation to have with the child. Feedback at this age is mostly verbal and relational, and a machine has almost nothing to offer it. If you find a reception or first-grade class receiving AI-written comments, ask about it, because the value of feedback at this age is that it came from the person who watched them do it.
8 to 12. Children this age take marks personally and literally. If a comment reads as generic, they notice, and what they conclude is that the teacher did not care rather than that a tool was used. This is the age where a bland comment does real damage, which is an argument for better feedback, not necessarily for less technology.
13 to 15. Old enough to raise it themselves, and many do, usually as "sir just used ChatGPT on that." Take it seriously without escalating. Ask what made them think so, then ask whether the feedback told them anything useful about their actual work. That second question is the one that matters. You could say: "Maybe they did. Did it tell you anything you can use?" A teenager who learns that the answer to a frustrating comment is to work out what is in it, rather than who wrote it, is being taught something worth knowing.
16 to 18. Now it touches coursework, predicted grades and applications, where a mark carries weight beyond the classroom. If a grade matters for a qualification, your school and exam board have formal rules that sit above any of this. Use the appeals or review process the school already has rather than an informal conversation, and ask what the mark rests on. If the dispute runs the other way and your child has been accused of using AI, that is a different process with different rules.
What no policy can guarantee
An honest list, because you should not leave this page more confident than the evidence allows.
You cannot tell from a piece of feedback whether a tool helped write it. There is no reliable check, and the detectors sold for student work are not accurate enough to use on teachers either. The same limits that make AI detection unreliable for your child's writing apply in reverse.
A school policy does not mean every teacher follows it, and no policy prevents a rushed judgment at eleven at night, with or without a chatbot.
Knowing a tool was used does not tell you the mark was wrong. Knowing no tool was used does not tell you it was right.
And three things take precedence over everything on this page. If you are formally disputing a grade, the school's own review process is the route, not a conversation about AI. If the qualification is external, the exam board's rules govern, not the school's preference and not your view of it. And if what is actually happening is that your child has lost confidence in a teacher, that is a pastoral conversation with the school, and it will go better if it does not arrive as a technology complaint.
Where this leaves you
Most teachers are not using AI to grade your child's work. Some are using it to help write the words, which is defensible and usually invisible. Almost none of them have been told where the line is, and that, rather than the technology, is the thing worth raising.
If you want one small thing to do tonight: find the last piece of written feedback your child brought home and read it for whether it refers to something only a reader of that particular piece could know. A specific sentence, a recurring error, an argument they made. That is not an AI test, because a good tool prompted well can produce specifics too. It is a better question than the AI one: did somebody actually read this?
Then ask the question rather than guessing at the answer. Schools respond well to a parent who asks what the position is. They respond badly to being accused of something nobody has defined.
The next step
The gap this article keeps running into is not a policy gap, it is a training one: teachers using a tool nobody has given them a rule for. If your child's school has not worked out where it stands on this, training school staff to use AI properly is the work I do with schools, and it is worth sending to your head of year or principal.
I also write a Sunday newsletter that keeps more than 44,000 teachers and parents current on this without the panic. Parents are very welcome.
Dan Fitzpatrick is a former teacher and assistant headteacher who has trained more than 150,000 educators, and now advises schools and governments on AI. More about Dan.
