Skip to main content

Proof of authorship

Why AI detectors flag human writing, and what actually holds up

If your genuinely original work got flagged, you are not an edge case. Detector false positives are structural, not bugs, and they will not be fixed by better detectors. Understanding why changes how you defend yourself.

Detectors measure typicality, not origin

AI-text detectors estimate how statistically predictable your prose is. But clear writing is predictable by design: students taught the five-paragraph essay, non-native speakers using learned patterns, technical writers following conventions, all produce exactly the regularity detectors flag. The tools measure conformity to expected patterns, and good writing often conforms.

Meanwhile, models improve at producing unpredictable, human-textured prose every year. The signal detectors depend on is evaporating while the punishment for a false flag stays constant.

The arms race only gets worse

Every detector improvement is training data for the next generation of models. This is a race where the fugitive builds the radar. Institutions are slowly recognizing this, and several detector vendors now publish disclaimers that scores should not be used as sole evidence of misconduct.

That is why we deliberately do not build AI detection. Guessing origin from text alone is a losing fight; recording origin as it happens is not.

What holds up instead: records over guesses

The durable answer to did you write this is a drafting record: the work, growing, timestamped, in increments no one would bother to fake. Version history is the accessible version of this. An independently sealed checkpoint record is the strong version: verifiable by the accuser directly, no trust in you required.

Against a probability score, a verifiable record of process is a categorically better kind of evidence. One is a guess about your text; the other is a fact about your time.

Build the record before anyone accuses you

Firsthand seals verifiable checkpoints of your work as you write: your words never leave your browser, only a fingerprint is sealed, and anyone can verify the timeline without trusting you or us. One document free, no card needed.

Protect your work, free

More on proving authorship