Quick answer: Yes – but its payoff moved. Google restricted FAQ rich results (the expandable SERP dropdowns) to well-known authoritative sites back in 2023, so most sites no longer get the visual treatment. The schema itself, however, still works hard: it hands AI engines clean question-answer pairs for extraction, feeds People Also Ask matching, and structures content for voice responses. In the AEO era, FAQ schema is extraction infrastructure, not a SERP decoration.
What changed and what did not
Changed: the visible FAQ dropdowns in Google results became rare, ending the era of marking up everything for SERP real estate. Unchanged: FAQPage markup remains valid, crawled and parsed – and answer engines consume it. When an AI engine needs a crisp answer to a question you have marked up, the schema serves it a labelled, self-contained unit – exactly the shape extraction wants (the AEO writing rules).
Where FAQ schema earns its keep in 2026
AI answer extraction: marked Q&A pairs are unambiguous evidence blocks for RAG pipelines. PAA alignment: questions phrased like real queries surface in People Also Ask matching. Voice: assistants read concise marked answers well. Long-tail coverage: a well-built FAQ block answers a dozen conversational prompt variants one page would otherwise miss.
Doing it right
Use real questions users actually ask – harvested from search data, prompt research and support tickets, not invented filler. Answer in 40-80 self-contained words. Keep visible content and markup identical. Place FAQs on the topically relevant page rather than one orphan FAQ dump. And do not mark up promotional non-questions – engines learned to ignore those years ago.
Testing the payoff
After adding genuine FAQ blocks, watch: PAA appearances for those questions, AI answers to the same phrasings across engines, and long-tail impressions in Search Console. Prompt-level movement is trackable in IndexGraph.ai – marked-up questions typically show retrieval-engine effects within weeks.
FAQ
Should I remove FAQ schema since rich results ended?
No – removal discards extraction value for zero gain. The markup costs nothing to keep.
How many FAQs per page?
Three to eight genuinely useful ones. Past that, split by subtopic – focused blocks extract better than encyclopedic dumps.
Adexorb Technologies builds FAQ layers from real prompt data as part of its AEO restructuring work.