Had three clients in the last month explicitly ask for ‘AI-detection-proof’ content as a deliverable spec. like they want guarantees the content passes their preferred detector. im a content strategist not a magician. detector results vary by tool, by week, by which model the detector last trained against. how are other freelancers handling this in your scopes of work
ive been navigating this for the last 6 months. my SOW language now reads: ‘deliverables will pass [client’s specified detector] at 80%+ human probability on first submission. if a different detector or future detector update flags content, additional revisions are billed at hourly rate.’ this protects you against the moving target. ive had 2 clients try to relitigate completed work months later when they switched detectors. having the SOW language meant i could quote them for the revisions and they backed down. for the actual passing-the-detector work: write naturally, then if it doesnt pass i run drafts through Walter Writes to restructure phrasing. its become a normal part of my workflow.
agree with the SOW language. one more thing i added: i specify the detector in the scope, including its version. ‘will pass GPTZero v3.2’ versus just ‘GPTZero’. when they update their algorithm im not on the hook to redeliver months of work.
Adopting this framing immediately. I lost two days last week on revisions for content that passed on delivery but flagged after a client switched to a different detector. Putting detector specification in every scope going forward.
@RustyCircuitX this is exactly the framing i needed. drafting updated SOW template this week. thanks for the detailed answer.