FAQ Blocks AI Answers Actually Cite: A Plain Owner Fix

Google removed the FAQ rich result on May 7, 2026, and a lot of owners read that as a funeral. It was not. The flashy accordion in search results is gone, but the question-and-answer block underneath it never stopped working. An engine does not cite a web page because it has a dropdown. It cites a page because that page already looks like an answer, and a clean question answered in plain words is the clearest answer shape there is.

Key Takeaways

• Google retired the FAQ rich result on May 7, 2026, but kept the FAQPage schema type alive.
• A question-and-answer block wins because its shape matches how AI engines assemble and assert answers.
• 34% of U.S. adults have used ChatGPT, so the buyer question your FAQ answers is a question your lead pool is already asking.
• The fix is four moves: a real question, a direct answer, the schema mark, and a page an AI crawler can reach.

TL;DR

The FAQ rich result is dead, but the FAQ block is not. What changed is cosmetic: the expandable box in Google results is gone. What did not change is the core mechanic: engines lift answers that are short, direct, and written under a question. The block still earns citations when it does four things. Ask the question your buyer actually types, answer it in three sentences or fewer, mark it with schema a machine can parse, and make sure an AI crawler is allowed to reach it.

What actually died, and what did not?

Here is the truth most owners missed. Google restricted FAQ rich results to government and health sites back in August 2023, then removed the FAQ rich result from search entirely on May 7, 2026, as noted in the Search Central documentation updates. The visual dropdown is gone for good.

What survived is the format. The FAQPage schema.org type is still a named, machine-readable way to mark a list of questions and answers. Google itself said, back in the 2023 announcement, that dropping the structured data was optional and that unused structured data causes no harm.

That distinction matters more than it sounds. Most owners conflated the box with the block, deleted both, and threw away the exact surface an answer engine reads. The block did not die. Only the box did.

Why does the question-and-answer shape still win?

An AI answer is assembled, not stored. When a buyer asks ChatGPT who to hire, the model pulls pages, finds the passage that answers the question, and paraphrases it back with a name attached. A page is more likely to be that passage when it is already shaped like the question, and that is the reason answer-first page structure works.

A question heading followed by a direct answer is the single most liftable structure on the web. There is no wind-up to strip. There is no ambiguity about what the page is claiming. The engine can lift the answer and cite the page in one move.

Pew Research found that 34% of U.S. adults have used ChatGPT, roughly double the share two years earlier. Your buyer is in that number. The question they type is the question your FAQ should answer, word for word.

The speakable property in schema.org is the clue most people skip. It lets a page flag the exact sections best suited to being spoken aloud. A voice assistant reads the flagged answer because the page told it which sentence is the answer. That is the whole game, reduced to one mark.

What makes an FAQ block get cited instead of skipped?

Four moves separate a block an engine lifts from a block it scrolls past. They are mechanical, and they compound.

1. Ask the real question. Use the exact words your buyer types, not the question you wish they asked. "How much does a new roof cost in Phoenix" beats "Roofing investment considerations." The first matches a real query. The second matches nothing.

2. Answer in three sentences or fewer. State the number or the fact first, then clarify. A direct answer is quotable. A paragraph of hedging is not. If you cannot say it plainly, you have not decided what you think.

3. Mark it cleanly. Wrap the block in FAQPage schema so the machine sees a list of pairings, not free text. This is the same mark that survived the rich-result shutdown and still helps any parsing system see your answer.

4. Let the crawler in. OpenAI's OAI-SearchBot decides whether your site shows up in ChatGPT search answers, and it only reaches pages your robots.txt admits. A perfect FAQ block behind a bot block is a block nobody ever reads.

None of these require a developer after the first setup. They require decisions: which questions, what answer, and a page that is reachable. This is the same shape that lets a local specialist get cited when the engine goes looking for a name.

Where does llms.txt fit into this?

The FAQ block is one answer on one page. An llms.txt file is a single location that points an agent to the handful of pages worth reading. Think of it as a table of contents written for a machine instead of a person.

The two work together. The llms.txt file tells the agent which page holds the answer. The FAQ block is the answer that page holds. One is the map, the other is the destination.

This matters for a small expert business because you do not need a thousand pages. You need three to five questions your buyers actually ask, answered so cleanly an engine can quote them without editing. The llms.txt file just makes sure the engine starts at the right page and not a buried contact form. Most of your market is in the same boat, and that is the opening: the SBA Office of Advocacy reports that 99.9% of American businesses are small, a pool of over 33 million firms that are almost all invisible to AI answers about their own trade. This is also the reason structured data still matters even when the visible boxes vanish.

Sources cited in this analysis?

The FAQ rich result removal date and the 2023 restriction to government and health sites come from Google Search Central documentation. The FAQPage and speakable type definitions come from schema.org. The 34% ChatGPT adoption figure comes from Pew Research Center's June 2025 survey. The OAI-SearchBot crawler behavior comes from OpenAI's crawler documentation. The llms.txt convention comes from the llms.txt v2 proposal.

Frequently Asked Questions

Do I still need FAQ schema if the rich result is dead?

Yes. The rich result box is gone, but the FAQPage schema type still marks your questions and answers as machine-readable. That parseable structure is what helps an answer engine recognize and lift your direct answer, even without the visible dropdown.

What is the difference between an FAQ rich result and an FAQ block?

The rich result was the expandable box Google showed in search results, and it is retired. The FAQ block is the questions and answers on your page itself, and it still works. One was a display, the other is content an engine reads.

How many FAQ entries should a page have?

Aim for three to five questions your buyers actually ask, each answered in three sentences or fewer. More than that dilutes focus, and fewer than three rarely covers the buyer's real decision. Write the questions in the buyer's exact words.

Do FAQ blocks help with voice search and AI answers?

Yes. A question followed by a direct answer is the cleanest shape for a machine to lift, and the speakable property can flag the exact sentences to read aloud. This is the same structure a voice assistant and an AI answer engine both extract.

Is this worth doing if I already rank on Google?

Yes, because ranking and citing are different contests. You can rank and still never be the named business in an answer. A clean, reachable FAQ block is built for the citation layer, not the results page, and it is exactly what feeds the answer your buyer reads next.