Rewrite headings as questions people ask
Few or none of your H2 and H3 headings are phrased as questions. That matters more than it sounds, because AI assistants don’t search with your customer’s original words. They rewrite every prompt into a set of cleaner sub-queries first, and a heading that literally matches one of those sub-queries is the strongest relevance signal a page can send. “Our Pricing” matches nothing. “How much does teeth whitening cost?” matches the exact string an engine just generated.
Why do question headings win in AI search?
Because retrieval is matching, and the machine tells us what it matches against. Google’s own AI Mode documentation describes the “query fan-out technique”: one question becomes many related searches issued simultaneously. Ahrefs’ 1.4 million prompt study found the same mechanic inside ChatGPT, with cited pages showing higher title relevance to the generated sub-queries than to the user’s original prompt.
Headings work like titles at the passage level. When your H2 mirrors a sub-query, your section is the passage that gets pulled.
Which headings should become questions?
Your key sections, and at least two per page, which is where our check draws its pass line. Don’t force it everywhere; a page where every single heading interrogates the reader sounds like a quiz. Convert the label headings that carry your money content:
- “Pricing” becomes “How much does it cost?”
- “Service Area” becomes “Where do you offer service?”
- “Our Process” becomes “What happens after you book?”
- “Warranty” becomes “What does the warranty cover?”
Brand and navigation headings can stay labels. The pattern you’re after is a page that reads like a good FAQ without formatting itself as one.
Where do the right questions come from?
Your customers, not your imagination. Mine your inbox and sales calls for the phrasing people actually use. Check the “People also ask” box on Google for your topic, since those are literal harvested queries. Ask whoever answers your phone which five questions they field every week.
Real phrasing beats invented phrasing for a mechanical reason: the fan-out queries are generated from how real people talk, so headings written in that register match more often. One practical upgrade from the citation research: engines frequently append modifiers like “best,” “comparison,” and the current year to their sub-queries, so headings and titles that naturally carry those words pick up extra matches.
What goes under the question?
The answer, immediately. A question heading followed by three paragraphs of wind-up wastes the match you just earned, because retrieval systems lift passages rather than pages. Open each section with a direct answer of under seventy words, then elaborate.
That structure is its own separate check in our scanner, and the two work as a pair: the question heading earns the match, and the front-loaded answer makes the lift easy. Our answer front-loading guide covers the second half, and the first-100-words guide covers the page-level version of the same idea.
Rewrite your key headings, redeploy, and rescan. This check carries 10 points, and it tends to move together with the two answer-shape checks around it, so a single writing pass often picks up all three.
See where your site stands. The free scan takes about fifteen seconds and shows every fix.
Run a free AI visibility scan