How Do Teachers Check for AI? What Schools Really Use in 2026

No single detector "catches" AI writing with certainty, and any teacher or tool claiming otherwise is overreaching. When UK teachers check submitted work for AI, they use several layers together: what they know of each student's writing, evidence from the drafting process, a conversation with the student, and, in more schools now, detection software treated as one indicator among several. Each layer is covered below, along with what exam boards expect.

The First Check Is Always the Teacher's Own Knowledge

Teachers read a lot of a student's writing over a year - classwork, homework, assessments, and everything in between. That builds an internal picture of vocabulary, sentence habits, typical errors, and how the student's ideas develop. The most common trigger for an AI conversation is not software at all; it is a piece of work that simply does not sound like the student who handed it in. A sudden jump in fluency, vocabulary the student has never used in class, or polished structure with none of their usual quirks - teachers notice, because comparing this piece against everything else the student has written is what they do all day.

Drafting Evidence and the Writing Process

The second layer is process evidence. Coursework and non-exam assessment (NEA) are produced over weeks, with drafts, plans, and supervised sessions along the way. When a final piece appears without a credible trail - no plan, no rough draft, a version history that shows a thousand words pasted in at once - that gap is itself evidence. This is why many departments now ask for work to be produced in supervised conditions or through documents with visible version history: the process protects students who wrote their own work just as much as it exposes those who did not.

The Conversation: The Most Reliable Check There Is

When a teacher suspects AI involvement, the next step in most schools is not an accusation - it is a conversation. A student who wrote their own essay can talk about it: why they chose that argument, what they meant in the third paragraph, what they would change. A student who submitted generated text usually cannot. UK exam boards, through JCQ's guidance on AI use in assessments, point in the same direction: teachers should use their knowledge of the student and the drafting process, and investigate concerns through dialogue and evidence rather than relying on any single tool's verdict.

Do Schools Have AI Detector Software?

Many do - but the way schools use it has matured. Detection tools analyse a piece of writing and return a likelihood score: an indication of how consistent the work is with the student's expected writing, not a verdict. Used well, that score is a prompt for the human steps above, never a substitute for them. False positives are real - strong formal writing, and writing by students working in English as an additional language, can trip naive detectors - which is why a score alone should never be the basis for an academic misconduct finding.

GradeOrbit's AI detection, which is part of its school accounts, is built on exactly that philosophy: it frames every result as an indicator to review - a conversation starter - and it compares a piece against the student's own previously marked work, which is a far fairer baseline than comparing against generic "human writing". A student with no earlier work on file is skipped rather than given a generic score.

How Do Exam Boards Check for AI?

Exam boards do not run a secret scanner over every script. For coursework and NEA, the responsibility sits first with the centre: teachers authenticate that submitted work is the student's own, and JCQ guidance requires schools to have malpractice procedures for AI misuse. Boards investigate when authentication is questioned - drawing on drafts, supervision records, and the school's evidence. In examined papers the question barely arises, because the work is produced under controlled conditions. The practical upshot for students: the person most likely to notice AI-written work is not an algorithm in an exam board office - it is the teacher who has read everything else you have written this year.

The Time This Takes

For teachers, the catch is workload: comparing styles and holding follow-up conversations takes time that is already short. GradeOrbit is one tool that addresses both parts. Its marking assistant drafts marks against your scheme for you to review, and on school accounts its detection tool gives an indicator to look into when something feels off, framed for professional judgment rather than accusation. You can see how school AI detection works before deciding whether it suits your department.