How Teachers Detect AI in Summer Bridging Work

Every summer, departments across the country set bridging work: the transition tasks that incoming Year 12 students complete between their GCSE results and the first week of A-Level study. It is also, quietly, the single easiest place for undeclared AI use to slip through unnoticed - which is exactly why learning to detect AI in summer bridging work matters before that first folder of completed tasks lands on your desk in September. Six unsupervised weeks, no classroom, no draft stages, and a student who has not yet met their teacher is a context almost designed to make AI-generated writing invisible.

This post is for the teacher opening that pile of transition work in early September, not the one setting it in July. The useful question is not "how do I stop students using AI over the summer?" - you largely cannot - but "how do I read what comes back fairly, and what do I do when something looks generated?"

Why Bridging Work Is Uniquely Vulnerable

Most of the signals teachers rely on to sense whether a piece of work is genuine are missing in a transition task. You have no prior writing from the student in your subject to compare against - they were at a different school, or in a different class, only weeks ago. There were no in-class drafts, no planning stages you witnessed, no lesson where you watched them wrestle with the idea. The work arrives fully formed, from a student whose voice you have never read before.

The tasks themselves often invite polished, self-contained answers: "read this extract and write a 600-word analysis", "summarise the key causes of the First World War", "explain how enzymes work". These are precisely the prompts a large language model handles fluently. A student who would normally produce solid grade-6 GCSE writing can submit something that reads like a confident A-Level response, and you have no baseline that tells you the jump is unusual - because for you, this is the baseline.

What a Likelihood Score Adds - and What It Doesn't

This is where a detection score earns its place. An AI detection tool returns a probabilistic likelihood score, expressed as a percentage, reflecting how closely the linguistic patterns in the text resemble those typically produced by AI models. For bridging work, where your own intuition has nothing to anchor to, that external signal is genuinely useful: it gives you a starting data point you would otherwise lack entirely.

But the score is a starting point, not a verdict. A high likelihood score on a transition task does not prove a student generated the work, and it carries an extra risk you should hold in mind: incoming students are an unknown population. Some are EAL students whose carefully formal register reads as machine-like to a detector. Some have spent the summer reading widely and writing in a deliberately academic voice because they think that is what sixth form expects. The score tells you a piece is worth a second look; it does not tell you what happened. For the full picture of how to read the number itself, our guide to interpreting AI detection likelihood scores goes deeper.

A Fair Process When You Have No Baseline

Because you cannot compare a transition task against the student's earlier work in your subject, the fair process shifts slightly. The first move is to build a baseline quickly rather than reach for it retrospectively. A short, supervised diagnostic task in the first week - a timed paragraph written in class on a similar topic - gives you a sample of the student's genuine voice under conditions where AI was not available. Held next to the summer work, a large gap in fluency, structure, or vocabulary range is far more telling than a likelihood score on its own.

The second move is conversation, not accusation. If a bridging task flags a high score and the in-class diagnostic looks very different, talk to the student about the work itself before mentioning any tool. Ask them to walk you through how they approached it, what they found difficult, which source they leaned on. A student who genuinely wrote it can usually discuss their own thinking; a student who generated it tends to repeat surface points and stumble on specifics. The principles here mirror those in our piece on handling a high AI likelihood score fairly.

Setting Expectations for the Year Ahead

The most valuable outcome of detecting AI in bridging work is rarely a sanction - it is a conversation that sets the tone for two years of sixth form. September is the moment to be explicit about where AI sits in your subject: that it is a legitimate tool for some things (clarifying a concept, checking understanding) and not for others (generating the analysis you are meant to produce yourself), and that the line is the same whether the work is set in July or in January.

Framing the bridging task as the first data point in an honest working relationship - rather than a trap - also makes detection more effective over the rest of the year. Once you have a few weeks of genuine, supervised writing from each student, your own judgment becomes the primary signal and the detection tool becomes a backstop for the pieces that surprise you. A transition task flagged in good faith and discussed openly teaches the student that the work is read carefully, which is itself a deterrent for the coursework that matters far more later. If those later pieces are heading to a moderator, our guide to using AI detection as professional evidence covers how to document it.

How GradeOrbit Helps with Transition Work

GradeOrbit's AI detection is included on school accounts, and it works differently from a generic detector: it compares each new piece with that pupil's own earlier work that the school has already marked in GradeOrbit, and returns a score from 0 to 100 for how far the new piece departs from their own writing, with a Low, Medium or High confidence label, the signals it noticed, and a plain-English explanation. That has an honest consequence for bridging work. A student who stayed on from your own Year 11, with GCSE work already marked in GradeOrbit, can be checked straight away. A student joining from another school has no earlier work on file, so they are skipped rather than given a generic score, and you are told. For them, the supervised diagnostic task above is your baseline - and once you have marked a few pieces of their work in GradeOrbit, detection has something real to compare against for the rest of the year.

Detection is unlimited and included in every school teacher seat, and you can run it on a whole class at once: upload scans, photos or PDFs, and bulk upload splits the pages by pages per student and matches the name on each cover page to your class list. Pupils' work is held under the school's Data Processing Agreement for its chosen retention window, scans are removed when a run archives, and work is never used to train AI models. Individual teacher accounts don't include detection; there, work uploaded for marking is processed and discarded, personal details are redacted before processing, and students are referred to anonymously as "Student 1", "Student 2", and so on. The score lands in your professional judgment, where it belongs, and the audit trail lives in your own records.

Try GradeOrbit on Your Bridging Work This September

If you want a fair way to check transition work against each student's own earlier writing, and a baseline that builds as you mark their work through the year, GradeOrbit's school accounts include AI detection for every teacher, with no per-check charge. It isn't offered on individual teacher accounts, because a generic "does this read like AI?" score on a single piece isn't reliable enough to raise with a student.

See how AI detection works for schools and start a free school trial. The tool is built to support your professional judgment as you get to know a new cohort - not to replace it.