Building an editorial AI-use policy in an afternoon

You don’t need a legal team to write this. You need four decisions and a place to put them.

Last reviewed 3 October 2026

Four decisions, not forty

1. What’s allowed without disclosure — research, brainstorming, fact-checking.
2. What requires disclosure — drafting assistance, editing passes.
3. What isn’t allowed at all, if anything.
4. What happens if a piece is flagged after submission — a conversation first, per using a detector as a second opinion, not an automatic rejection.

Put it in the submission guidelines, not a separate document

Contributors read submission guidelines. They don’t read a linked policy page they have to click through to find. Say the actual rule where the actual deadline is.

Say what a score does and doesn’t mean, if you check

If you run submissions through a detector, tell contributors that a mid-range score isn’t a rejection on its own — see the six signals in editorial terms for language you can borrow. Contributors who understand the process file fewer disputes than contributors who find out about it after being flagged.

Revisit it in six months

This is a fast-moving area; a policy written today will need a look again once tools and norms shift. Put a date on it and actually check back.

Keep reading