Originality.ai and false positives

A detector marketed to publishers and SEO teams, sold on accuracy and used to screen content at scale.

What it claims

The vendor acknowledges false positives are a known risk. Independent testing found it produced the highest false positive rate of the three leading detectors tested, flagging 3.8% of genuine human passages.

Where it fails

It is built for SEO content, which is exactly the structured, keyword-shaped prose that trips detectors. That combination produces accusations against writers who did the work.

How to respond

Ask for the raw score across the whole document rather than a verdict, and note the measured false positive rate. A single flagged paragraph in a long piece is not a finding about the piece.

Worth knowing

It also sells a humaniser that rewrites AI text to evade detectors. The same company profits from both sides of the same dispute, which is worth knowing when its output is used against you.