Editors are often the last people to read a piece before it carries a publication's name. Winston AI flags sentences that read as machine-written in freelance drafts and contributed articles, so you know where to push back. This page covers contributor guidelines, raising concerns with writers and handling drafts that are only partly generated.
Put AI in your contributor guidelines
The simplest way to avoid disputes is to say what you accept before anyone files. One clear paragraph in your guidelines turns an awkward conversation into a reference to an agreed rule.
- State whether AI tools may be used for research, outlines, grammar or drafting.
- Ask contributors to disclose any tool use when they file.
- Explain that drafts may be checked, and that a flag leads to questions, not automatic rejection.
- Require sources or notes for factual claims, which protects you however a draft was written.
Partly generated drafts are the common case
A fully generated piece is less common than a draft where the writer used a tool for a paragraph or two: an intro, a summary, a transition. Sentence-level highlights show exactly which paragraphs to query. Sometimes the fix is a quick rewrite. Sometimes the flagged part is where the reporting should have been, and that's the conversation to have.
Expect pushback now and then, especially from writers who did the work themselves. Lead with specific questions, not the score, and accept notes, recordings or links as answers. If the reporting checks out, the piece goes ahead. The highlight did its job by getting you to look.
If you're setting rules for a whole site rather than one desk, our page for publishers covers site-wide standards and archive audits.