This is a style linter, not an authorship detector. It finds habits language models exhibit more often than most human writers — nothing more. Seven commercial detectors once flagged 61% of human-written TOEFL essays as machine-written (Liang et al., Patterns, 2023), against 5% for US eighth-graders, because careful formal prose in a second language looks exactly like what a model produces. Use this on your own drafts. Never use it as evidence against a person.