How it works
Paste your headings (one per line, with optional ## markdown prefixes for H2/H3) and the tool scores each one on six factors: question shape, word count, extraction-friendliness, heading level balance, starter verb and overall answer pattern. The aggregate score gives you a grade (A-D).
Why heading shape is the #1 structural lever
AI assistants passage-extract content at the section level, not the page level. When a heading is already shaped like a question ("How do I measure GEO ROI?") the assistant gets a matched query-answer pair for free - zero re-parsing, zero extraction ambiguity. The same content with a dead heading ("GEO metrics") still gets indexed, but is 3-5x less likely to be pulled into an answer card. The full format playbook is in our content-to-cite guide.
The fix patterns
- Rewrite dead noun-phrase headings as questions: "GEO metrics" → "How do I measure GEO ROI?"
- Keep headings between 6-12 words - long enough to carry context, short enough to not bury the answer.
- Front-load the keyword: the first 2-3 words of the heading carry the most extraction weight.
- Match headings to the FAQ schema mainEntity pairs exactly - this is what gets you pulled into AI Overviews.