AI detector flagged your essay? What to do next

A detector score is a model’s judgment about patterns in text. It is not a recording of who typed the words or how the draft was made.

The accusation can feel settled when a report gives a percentage. But even Turnitin says its model can misidentify human writing and should not be the sole basis for adverse action against a student. Oxford’s AI Competency Centre says the university does not endorse digital AI detectors for academic decisions.

If the result concerns your work, keep your outline, notes, source list and draft/version history. Ask which passages were flagged, what tool produced the report, and which written course policy applies. Explain your process plainly and share the writing record you already have; no single draft or detector score proves authorship by itself. If you used AI, describe that use honestly against the assignment rules. Do not rewrite your work to “fool” a detector or invent a paper trail.

Recent research shows why a single score needs context. A September 2026 study of Turkish academic writing tested several detector systems under specific, held-out conditions. One deployed system falsely flagged 35% of the human-written texts in that benchmark. That is one system, one study setup and one language context—not a false-positive rate for every detector, school, language or current version. A 2023 study also found that several then-available detectors disproportionately misclassified a small set of non-native English essays; those older results should not be treated as a current test of every tool.

Before pasting unpublished coursework into a free checker, read its privacy and retention terms. Different tools have different policies, and an extra score may not resolve the dispute. The useful question is not just “What percentage?” but “What independent evidence supports this conclusion, and what process does the institution require?”

Research results depend on the detectors, sample, language and conditions tested; this is not legal advice or a verdict on any individual case.

Sources: Turnitin; University of Oxford; Tüker et al. (2026); Liang et al. (2023).