Define real review conditions
The public protocol requires separate human-only, AI-only, mixed-authorship, edited, translated, short-form, and domain-specific samples before any result can be interpreted.
Resources
A concise summary of WriteGo's public AI-detection evaluation protocol. Auditable performance results are pending; the page covers dataset scope, false-positive reporting, edited drafts, multilingual samples, and review limits.
Open core guideThe public protocol requires separate human-only, AI-only, mixed-authorship, edited, translated, short-form, and domain-specific samples before any result can be interpreted.
Any future report must separate false positives and other errors by language, document length, template use, and writing context instead of relying on one headline accuracy figure.
WriteGo has not yet published an auditable performance report. Numeric results will be added only with dataset definitions, a versioned procedure, subgroup results, error analysis, and reproducible materials.
It defines sample categories, model and procedure versioning, editing conditions, language coverage, subgroup error reporting, confidence bands, and limits on how future results may be used.
Not yet. The protocol is public, but the supporting dataset definition, versioned procedure, subgroup results, and reproducible materials are still pending. Individual decisions always require passage evidence, context, policy, and human judgment.