A score is not an authorship certificate
Atlas displays the provider’s score. A high value indicates stronger AI-like patterns in the analyzed input, but it does not establish who created it. It is not a measured percentage of AI-written words.
Do not treat the result as proof. Review drafts, source material and revision history when they are available and shared with permission.
Why results can disagree
Models and thresholds differ. Short passages, formulaic writing, translation and editing can affect results. A low score does not establish human origin, and a high score can be a false positive.
What the video report covers
Atlas’s video beta shows the highest score among returned frames. It does not inspect every moment of a clip and does not analyze its audio. An unflagged sampled frame cannot validate the entire video.