Ask Claude or ChatGPT to break a head term into subtopics, search intent, and 'people also ask' questions — paste in real search-volume data so it doesn't hallucinate numbers.

Answer-first structure: H1/H2/H3 headings plus a one-line summary of what each section needs to cover.

Don't generate the whole post in one shot — feed it a voice sample and the 3 specific facts each section must include. One-shot drafts read generic.

Every number, name, law, and statistic against the primary source — this is the step people skip and the one that actually matters. LLMs invent statistics that sound completely plausible.

Title tag, meta description, internal-link anchors, FAQ schema, and natural keyword placement — not keyword-stuffed, just present where it reads naturally.

Strip the em-dashes and other AI writing tics, vary sentence length, add one concrete example or first-person line, then read the whole thing aloud — you'll hear what still sounds robotic.

A real author bio, an original screenshot, an actual quote — then publish and check the live URL's rendered title, H1, and meta tags actually match what you wrote.