AI detection FAQ
What is a good AI detection score?
Quick answer
There is no universal cutoff. A detection score estimates how AI-like the writing appears; it should guide review, not act as a pass/fail threshold, because the right interpretation depends on context and stakes.
A low score does not guarantee text is human-written, and a high score does not prove misconduct. Scores reflect probability and confidence, which vary with document length, editing, language, and topic.
Rather than chasing a single number, look at which passages are flagged, how strong the signal is, and whether the result fits the document's draft history and sources. High-stakes decisions deserve more evidence than a borderline percentage.
Set policy around process, not just thresholds: define when disclosure is required, how reviewers handle uncertainty, and what follow-up (revision, conversation, or escalation) a flagged result triggers.
Check text with the AI detector
See explainable, passage-level evidence instead of a single score.
Open the AI detector