For teachers & students
AI essay checker
Check an essay for signs of AI writing without sending student work to a third-party server. Every sign is highlighted and explained — so a result can start a conversation instead of ending one.
Free · no sign-up · your file never leaves your device
Your text, with the signs marked
Hover a highlight to see which sign it matched. Invisible characters are shown as labelled boxes.
What detection can and can’t do in a classroom
AI detectors are statistical tools. They estimate how much a text resembles chatbot output, and they are wrong some of the time in both directions. Published research and the experience of many schools show the same pattern: detectors catch a lot of unedited AI text, miss text that has been paraphrased or rewritten by the student, and occasionally flag genuine work — more often for writers who use a formal, careful, formulaic style, which includes many non-native English speakers and students taught to follow strict essay templates.
That does not make detection useless, but it changes how to use it. A detector is good at telling you where to look. GPTTrace is designed for that: rather than a single score, it highlights the exact sentences that match each sign and explains why the sign is associated with AI writing.
Signs that matter most in essays
- Generic significance claims — “This pivotal moment left an indelible mark on history” with no specific evidence.
- Summaries that restate — a conclusion that repeats the introduction’s three points in the same order.
- Vague sources — “historians argue”, “studies have shown”, with no citation, or citations that don’t exist.
- Perfectly even structure — five paragraphs of nearly identical length, each opening with a transition word.
- Leftovers — “Certainly! Here is an essay on…”, Markdown symbols, or invisible characters from copy-pasting.
Invented or mismatched citations are often the most concrete evidence of all, and no detector checks them: look up two or three references.
Designing assignments that hold up
The most effective responses to AI in coursework are about the assignment rather than the detector: writing in class, staged submissions with drafts, personal reflection tied to class discussion, oral follow-ups, and topics specific to your course materials. Detection then becomes a backstop rather than the main line of defence — the role it is actually suited for.
How accurate is it? Our measured numbers
We test GPTTrace on labelled text samples and publish the results, including where it does badly. It is tuned to keep false accusations rare, so it misses some AI content rather than flag real work.
standard text check: AUC 0.795 (cross-validated)
2,467 labelled samples (1,258 AI, 1,209 human), run 2026-10-08. At the “Likely AI” line it caught 34% of AI samples and wrongly flagged 5% of human ones.
| Source | Truth | Samples | Result at “Likely AI” |
|---|---|---|---|
| pd-literature | Human | 306 | 1% wrongly flagged |
| human-pmc-esl | Human | 183 | 5% wrongly flagged |
| human-wikipedia-pre2022 | Human | 155 | 8% wrongly flagged |
text check with deep scan: AUC 0.91 (cross-validated)
2,467 labelled samples (1,258 AI, 1,209 human), run 2026-10-08. At the “Likely AI” line it caught 56% of AI samples and wrongly flagged 5% of human ones.
| Source | Truth | Samples | Result at “Likely AI” |
|---|---|---|---|
| pd-literature | Human | 306 | 0% wrongly flagged |
| human-pmc-esl | Human | 183 | 1% wrongly flagged |
| human-wikipedia-pre2022 | Human | 155 | 1% wrongly flagged |
Data sources and method: methodology & accuracy.
A fair process when an essay looks AI-written
- Run the check and read the highlighted passages, not just the percentage.
- Compare with the student’s earlier work written in class or before AI tools were common.
- Ask for process evidence: notes, outline, drafts, document version history, sources.
- Talk with the student about the argument and the sources. Someone who wrote an essay can explain and extend it.
- Decide on all the evidence together. Never on a detector score alone.