AI in education, as a phrase, has run ahead of the evidence. In 2026 the usage numbers are clear, the outcome numbers are not, and the official guidance is more cautious than most of the coverage suggests. This piece is the reference version: what UK and US students are using, what schools policy says on both sides of the Atlantic, how exam boards handle AI, what teachers are doing, what parents should be asking, and what's likely to change in the year ahead.
Every number here is tied to a named primary source with a date. Where the evidence is thin, that's flagged. The goal is to be quotable in the way official reports are quotable, not in the way headlines are.
Where things stand in 2026: AI use has grown rapidly among students, but evidence on educational outcomes remains limited (Ofsted, June 2025). Official policy is cautiously supportive, not enthusiastic. That gap between adoption and evidence is where most of the confusion sits.
Where we are in 2026: the usage picture
Four primary surveys anchor the numbers on how many students and teachers are using AI. Their dates matter, because the figures move fast.
On UK teenagers, Ofcom's 2026 Children and Parents: Media Use and Attitudes report (21 May 2026) found that 56% of UK 8 to 17 year olds had used AI, rising to two-thirds of 16 to 17 year olds, and half of those users (51%) used it for schoolwork or homework. Ofcom also found teenage AI users split on whether they would trust an AI-written news article as much as a human-written one. There isn't a UK survey that isolates KS4 and KS5 revision-specific use, which is a real gap given how much of the discourse assumes one.
On UK undergraduates, HEPI's Student Generative AI Survey 2026 (fieldwork December 2025, published 12 March 2026, n=1,054) found that 95% used AI in some form in 2025/26, and 94% used generative AI for assessed work. 12% pasted AI-generated text directly into assessed work, up from 8% the year before and 3% in 2024, the number that has grown fastest.
On US teens, Pew Research's How Teens Use and View AI (24 February 2026, n=1,458 teens 13 to 17) found that 54% had used chatbots for help with schoolwork, and 10% did all or most of their schoolwork with chatbot assistance. Common Sense Media's Teens in the AI Era (August 2026) put the schoolwork use figure at 70%. 63% of US teens who use AI say they get answers to schoolwork from it; 25% say they use those answers as-is, while others rewrite them (31%) or improve them with what they know (35%).
On US teachers, Walton Family Foundation and Gallup's 2025 research, Six Weeks a Year, found that 32% used AI at least weekly, saving an average of 5.9 hours per week (about six weeks over the school year). 40% still don't use it at all.
What UK schools policy says
The Department for Education's Generative Artificial Intelligence in Education paper (last updated 12 August 2025) is the UK source of record. Its position is cautiously supportive, not a mandate.
The DfE states that generative AI "cannot replace the judgement and deep subject knowledge of a human expert", and that pupil use should be "carefully considered and assessed". It also expects schools to comply with the age restrictions set by AI tools, which vary: some require users to be 13 or over with parental consent, while others, such as Claude, are 18+. That's a compliance point most parents and teachers underestimate.
Ofsted's June 2025 early-adopter research studied 21 schools and further education colleges that had introduced AI. Its headline findings are worth quoting directly: "robust and reliable evidence of AI's impact on educational outcomes is limited", with most research described as "explorative, short-term and in limited domains". Reducing workload, especially for lesson planning and resource creation, was one of the two main reasons providers introduced AI; the other was supporting pupils' learning directly. Ofsted also confirmed there is no stand-alone section of school inspection focused on AI at this stage.
What UK exam rules say
JCQ's AI Use in Assessments guidance (30 April 2025) is the binding document for coursework and non-exam assessment. It treats "copying or paraphrasing sections of AI-generated content so that the work submitted for assessment is no longer the student's own" as malpractice, and "failing to acknowledge use of AI tools when they have been used as a source of information" as malpractice. Where AI is permitted, students must reference the tool, its web address and the date, as in JCQ's example ("ChatGPT 3.5 (https://openai.com/blog/chatgpt/), 25/01/2025"), and keep a non-editable copy (such as a screenshot) of their prompts and the AI's responses.
Ofqual's malpractice statistics give the scale. Summer 2025 recorded 100 cases of AI-related plagiarism, up from 85 in 2024. That was three-quarters (75%) of all student plagiarism cases but only 2% of the 5,025 student malpractice cases overall. Mobile phone and communication device offences remained the biggest single category at 44.3%.
So AI cheating in UK exams is a real category, and it's being caught, but the scale is smaller than the discourse implies. That framing matters, because policy responses that treat AI misuse as an epidemic are working from the wrong prior.
What US frameworks say
The US doesn't have a single federal equivalent to the DfE's paper. K-12 policy is largely set at state and district level, and it varies. What's clearest at national level is the College Board's position for the AP programme.
College Board's AP Exam Terms and Conditions state: "No part of your Test answers, work, or submission ... may be generated in whole or in part using any AI tools, including but not limited to ChatGPT." That applies to every AP exam. Its separate AI guidance sets course-specific rules for AP Seminar, AP Research, AP Art and Design and AP Computer Science Principles that are more nuanced. AP Capstone allows generative AI as an optional aid for exploring topics and initial source searches, but "all performance tasks submitted in AP Seminar and AP Research must be the student's own work". AP Art and Design allows AI in the early artistic process but prohibits its use in the final submitted artwork, and requires disclosure in the AP Digital Portfolio.
The UK and US frameworks are not equivalent and shouldn't be compared as if they were: JCQ and College Board aren't interchangeable, nor are the DfE and the US Department of Education. Anyone using the two systems interchangeably in policy discussion is glossing over meaningful differences.
Ready to boost your grades?
Join 1M+ students who have used Cognito to ace their exams.
Get started for free!
What teachers are (and aren't) doing
The teacher picture in the US is the clearest, thanks to the Walton/Gallup data. 32% use AI at least weekly and save around six weeks a year. 40% don't use AI at all. That's a real split, and it maps onto why student experience of AI in class varies so much.
On the UK side, Ofsted's early adopter research found workload reduction (lesson planning, resource adaptation, feedback drafting) was the main reason schools introduced AI, and the area where the impact was most visible. That's consistent with the US picture: teachers are using AI as a productivity tool more than as a pedagogy tool.
Common Sense Media's US teen data flags the gap this leaves. 73% of teens had never talked to a teacher about what AI is or how it works. Only 27% said a teacher had ever discussed what AI is or how it works with them, and only 30% had discussed how to use AI safely. 35% either didn't understand their school's AI rules or weren't sure what they were. Even in schools where teachers are using AI themselves, most students aren't getting a clear conversation about what safe use looks like.
What parents should ask their child's school
The DfE guidance and the Ofsted research together suggest three questions worth putting to your child's school, all of which map onto policies schools are expected to be able to answer.
What's the school's policy on pupils using AI for homework and coursework? Given JCQ's rules apply nationally to GCSE and A-Level, and the DfE expects schools to have considered pupil use, there should be a documented answer. If there isn't, that's useful information in itself.
What's the school's policy on age-gates for AI tools? Minimum ages vary: ChatGPT requires users to be at least 13 with parental consent under 18, while Claude requires users to be 18+. The DfE expects schools to comply with these age restrictions.
How are teachers being trained on AI, and what are they using it for? The DfE published support materials for teachers in June 2025. If your child's school has adopted AI for teacher workload but hasn't done pupil-facing training, that gap is worth flagging.
What students should do
The primary evidence points at one clear pattern to avoid, and one clear set of habits to build.
The pattern to avoid is offloading. Common Sense Media's August 2026 finding that 63% of US teens who use AI get answers to schoolwork from it, and 25% say they use those answers as-is, is the one that matters here. If AI is doing the retrieval, the problem-solving, or the writing for you, you're not building the skills the exam measures. The tool isn't the risk; the pattern of use is.
The habits to build are the ones that use AI for scaffolding and keep the retrieval yours. Cognito has two guides that go deeper: Should you use AI for your homework? An honest guide walks through the decision tree, and How to use AI to study effectively (without cheating) covers the specific techniques. For UK GCSE and A-Level students, How AI is changing GCSE and A-Level revision has the JCQ-specific detail. For AP students, AI in AP exam prep: What students are really using in 2026 covers College Board policy in depth.
What's likely to change in 2026-27
Two things worth watching.
In the UK, the Curriculum and Assessment Review chaired by Professor Becky Francis CBE published its final report, "Building a world-class curriculum for all", on 5 November 2025. The government published its formal response the same day. The review recommended strengthening citizenship education, including media literacy. The new national curriculum is due to be published in 2027 for first teaching from September 2028. Cognito's full breakdown is at GCSE curriculum review 2026: What's on the table.
On exam rules, both JCQ and College Board review their AI guidance regularly; JCQ's most recent update (October 2025) mainly added support materials for schools. Further changes are possible, but none has been announced. The exam-day ban on AI-generated content isn't going anywhere.
The honest verdict
In 2026 the honest read is that AI is embedded in how students study, official guidance is cautiously supportive rather than enthusiastic, and the outcome evidence remains limited. Any headline claiming AI is either "transforming education" or "destroying learning" is running ahead of what the primary sources say.
For teachers, the workload use case is the strongest evidence-backed one. For parents, the biggest lever is understanding how your child uses AI, not whether they use it. For students, the risk isn't the tool; it's the pattern of using it to skip the thinking, and that's the framing that survives the next 12 months of policy debate.
