FAQ schema for rich results and AI

Google got pickier. Only questions that sit on the page. JSON in the footer that nobody sees: do not count on that any more. I have seen landings with forty hidden questions and three sentences of visible text. That used to be a trick. Now it is a risk.

Oddly, those same visible blocks still help AI. Question, short answer, ready to cite. Three to eight questions people actually type. Two to four sentences each. An accordion is fine if the text is in the HTML, not fetched after a click. “Read more” that pulls the rest via JavaScript does not count.

No ad in the answer. No “call us” as the only content. Then mark that same text as FAQPage. Not the other way around: schema first, “we will invent the questions later”. And not the same five questions on every URL. That is one page, five times.

If the copywriter changes the answers, the schema has to move with them. Otherwise they lie to each other. Generate the JSON-LD next to the HTML, not in a cache that lags a week. An old price in a FAQ answer is worse than no FAQ.

I put FAQs on the page that actually answers the question. Not on home “so we have them everywhere”. One sharp page beats five landings with the same three questions. And if Google later drops the rich result, the text is still useful to a person. That is the test.