Limits

When AI detectors get it wrong

Every detector makes two kinds of mistake. One of them lands on a person who wrote their own work.

01 · Two mistakes

False positives and false negatives

Says AI
Says human
Human wrote it
False positiveA person is accused of using AI.
CorrectHuman text passes.
AI wrote it
CorrectAI text is caught.
False negativeAI text passes as human.

Detectors trade one for the other: a stricter threshold catches more AI text and flags more people. Ask any vendor for both rates.

The public record

What the evidence shows

61.3%
of TOEFL essays by non-native English writers were classed as AI, averaged across seven detectors. Liang et al., 2023
26%
of AI text was caught by OpenAI's own classifier, which also flagged 9% of human text. It was retired in July 2023. OpenAI

02 · Who it hurts

Where false positives land

WriterWhy a detector misreads them
Second-language writersPlainer vocabulary and predictable phrasing look like model output to a detector. Liang et al., Patterns, 2023
Formulaic genresLab reports, legal summaries and product copy follow templates, so they score as predictable.
Short textsA few sentences give a detector little to measure; scores swing on small changes.
Heavily edited human textGrammar tools and rewrites smooth the irregular patterns a detector reads as human.

03 · Evidence, not proof

How much a score is worth

  1. Level 1A score aloneA number from one tool. Never enough to act on.
  2. Level 2Score plus passagesThe sentences it flagged, read by a person who knows the writer.
  3. Level 3Plus a second toolTwo detectors agree, after both passed your own test.
  4. Level 4Plus the writing processDrafts, version history and a conversation with the writer.

04 · What to do

When a score says AI

  1. Read the flagged passages, not the number

    Ask whether they sound like the writer you know. A detector that cannot show passages gives you nothing to check.

  2. Ask for the process

    Drafts, notes and version history (Google Docs and Word keep it) say more than any score.

  3. Talk before you decide

    Ask the writer to explain their choices. People who wrote a piece can talk about it.

  4. Know your tool's error rate

    Run the 20-text test on writing like the one in front of you. If it flags your own human texts, it will flag theirs.

Some institutions have switched AI detection off for these reasons, Vanderbilt University among them in 2023: its guidance.