Most AI content gets filtered out of YMYL SERPs because it reads like a confident summary with no source of truth behind it. The fix starts at the prompt. My structure gives the model a clear role, the named author's real expertise, the exact sources it must cite, the claims it is not allowed to make, and a requirement to flag anything uncertain. The output reads like a careful expert, not a content mill. What does your current YMYL drafting prompt force the model to do?