Your expertise. Traceable sources.
Policy version: original-expertise-en-v1. English-language pilot. These rules apply to uploaded documents, pasted text and method interviews.
Submit your own substantive expertise
Use original materials you created, such as your course notes, training slides and documented methods. Do not upload AI-generated courses, invented experience, copied books or another expert’s work as your own. Owning a file or buying a course does not establish authorship or redistribution rights.
Disclose AI assistance
Transcribing your own teaching or correcting spelling is allowed if the substantive expertise remains yours and you review the result. Disclose this assistance. AI-generated substantive explanations, methods, examples or complete training materials are not eligible. Do not rewrite AI content to evade detection.
Provide provenance and protect people
Describe when and how you created the material and what original files, version history or course evidence can support it. A statement or file timestamp alone is not proof. Keep client records, participant lists, personal data and confidential third-party information out of the knowledge source. Rights and privacy review remain required.
Understand the AI screening
We run a local English text detector before publication. Its AI suspicion score is a statistical signal, not a measured fraction of AI-written words or proof of authorship. Scores strictly above 50% block publication. A score of 50% or below is only one requirement; it does not grant publication or prove originality.
Missing, failed, stale or unassessable checks also block publication. Short text, unsupported languages and extraction problems may prevent assessment. We check all text segments within supported limits and use the highest segment score. Reviewers can request original evidence. A disputed result does not automatically label you dishonest; you may provide evidence for review, but the score block cannot be bypassed in this pilot.
Review every version
Changes invalidate the prior check. Each active source needs a current declaration and passing check, followed by rights review and expert quality acceptance. Documents are processed on the application server, are not sent to a third-party detection API and are not used to train detection models. Local processing does not replace a complete hosting and privacy assessment.