SIGNAL / 26

Audit an AI detector false positive

Evidence before accusation

Paste the exact passage that was flagged. The audit separates measurable writing patterns, formatting sensitivity, hidden Unicode and reader-quality findings from the detector's unsupported authorship claim.

Audit the flagged textWhat to collect

Runs on this device

Run a false-positive audit

Use the original text, before rewriting or reformatting it. The local report gives you reproducible measurements to compare with drafts, citations, version history and the detector's own documentation.

0 words · add text to inspect exact counts and hidden characters

What the local check returns

AI signals you can inspect—not a verdict you have to trust

The report explains what was measured and keeps three questions separate.

T

Writing patternsRepetition, sentence rhythm, vocabulary and punctuation

P

Page evidenceDeclared metadata, page structure and resource references

Q

Reader qualityLong sentences, repeated phrasing and hidden Unicode

No AI probability score. This release describes signals but does not claim who wrote the text.

Fair-review checklist

A score is a question, not a verdict

  1. Save the detector name, version, date, threshold and complete report.
  2. Compare the exact submitted bytes with the original draft and its revision history.
  3. Check whether translation, templates, short length, formatting or accessibility tools affected the input.
  4. Review citations and let the author explain the work before any consequential decision.

This audit does not calculate an AI probability. It documents observable text evidence and the limits that a fair review needs.

Why independent evidence matters

The University of Melbourne's staff guidance says a detector result alone is not sufficient evidence for a misconduct allegation.

Turnitin's own guide also warns that its model may misidentify human-written, AI-generated and AI-paraphrased text.

Read before using a result

What this cannot tell you