This audit has not run yet. The table below shows the shape the published run will take — 9 detectors across 4 writer cohorts, with confidence intervals and per-domain splits. The detectors stay anonymous and the figures are placeholders until there are measurements to put in their place.
placeholder data · sort by clicking a column
Detector H6.5%30.5%35.3%23.0%28.8pp
Detector F4.4%24.5%30.2%17.5%25.8pp
Detector C3.8%21.6%30.1%16.4%26.3pp
Detector B2.2%19.7%24.1%12.2%21.9pp
Detector I3.0%17.0%21.1%12.2%18.1pp
Detector G3.0%16.9%25.2%13.5%22.2pp
Detector D2.5%15.2%21.4%11.3%18.9pp
Detector E2.0%13.5%20.4%10.8%18.4pp
Detector A1.3%11.7%18.7%14.8%17.4pp
9 detectors · all domains · figures illustrative for this build · methodology and corpora will be published with the first run
Widest cohort gap in the placeholder set
28.8pp
Detector H flags 6.5% of native-English writing and 35.3% of writing drafted with assistive tools. Placeholder figures, shown to illustrate the shape of the finding.
Methodology
Nothing here has been measured yet. This page is the published plan for an audit that has not run. When it does, corpora will be donated, consent-verified human writing with documented provenance; every run will be scripted and reproducible; cohort labels will be self-reported at donation time and never inferred. Detectors will be named once there are real numbers beside them. Until then the table below is placeholder shape, not findings.
Native English · n = 5,200 — First-language English writers, no assistive tools declared.
ESL writers · n = 3,400 — English as a second or later language, self-reported with proficiency band.
Assistive tech · n = 2,100 — Dictation, word prediction, or grammar assistance used during drafting.
Autistic writers · n = 1,700 — Self-identified autistic writers, diagnosis not required for inclusion.
Reporting on a detector case? Newsrooms and accredited researchers get the full runs and the raw data free when the audit publishes. Tell us what you need and we will build the export.