Everyone has a private position on this and very few have a public one. Some teams are publishing AI-assisted content at a volume they would not have thought possible two years ago, and seeing it perform. Others have decided that a single obviously generated paragraph costs more credibility than a quarter of traffic is worth. Both are looking at the same audience. This panel puts those positions next to each other and works out where the line actually falls, treating blog, documentation and social separately, because developer tolerance varies enormously across the three. Key learnings: - Where practitioners have found the tolerance limit in each format, and what happened when they crossed it - Which editorial checks are worth keeping as volume goes up, and which are theatre - How to tell whether your audience has noticed, before your engagement numbers tell you