rant incoming because this just became a real problem at my agency
we just got a policy update saying all external written deliverables need to include an “AI assisted: yes/no” declaration on the cover sheet. which is fine, i get it. but what about all the writing we use AI to help draft that we then heavily edit? where’s the line?
more practically: can humanise AI be detected by the tools agencies and clients are now using? because if we’re disclosing anyway it changes how i think about humanizers. if they’re not reliable i might as well just spend that time editing manually.
talked to a colleague who runs AI checks on vendor deliverables and apparently a leading AI detector catches humanized content pretty consistently. but GPTZero doesn’t as much. so your “clean” output depends entirely on which detector the other side is using.
genuinely wondering if humanizer tools are still worth it in this environment or if the transparency play is just better