Google's FAQ rich result is gone for every site, and what FAQPage markup still does

What Google switched off

The FAQ rich result was the list of expandable questions Google showed under a search listing. In 2023 Google limited it to well-known government and health sites. On 8 May 2026 it said "This feature will no longer appear in Google Search starting May 7, 2026." On 15 June Google removed the documentation as well. Both entries are in Google's Search Central changelog.

So no site gets the FAQ result any more, however well its markup is written.

Whether to take the markup off

Google answered that in its 2023 post on the HowTo and FAQ changes, when it first narrowed the result. "Structured data that's not being used does not cause problems for Search, but also has no visible effects in Google Search," it wrote, and there is "no need to proactively remove it."

In Google, then, FAQPage markup costs you nothing and earns you nothing. The real question is whether anything else reads it.

Who else reads it

Microsoft says it does. Fabrice Canel of Microsoft Bing confirmed on stage at SMX Munich in March 2025 that schema markup helps Microsoft's large language models understand content, as Search Engine Land reported under the headline "Microsoft Bing/Copilot use schema for its LLMs".

For ChatGPT and Perplexity the evidence is a test. In February 2026 Mark Williams-Cook put a made-up company's address inside JSON-LD that was not even valid schema, and left it off the visible page. Both ChatGPT and Perplexity found the address. Search Engine Roundtable covered the test, and the result suggests both read the markup as more text on the page. Whether the schema was valid made no difference.

What nobody has shown

No source I found shows an AI assistant recommending or citing a page because it carries FAQ markup. The test above shows that assistants read the words inside the markup. It does not show that the markup makes them pick you. Anyone selling FAQ schema as a way into ChatGPT's answers is going further than the evidence does.

That leaves a modest case for keeping it. A question and its answer, marked as such, give a program one less thing to work out, Bing has said it uses that, and Google has said it does no harm.

What AI Boost writes

AI Boost for Joomla writes one FAQPage block per page, with at most 20 questions and no question twice. It takes them from three places.

• Blocks you mark. On the Structured Data screen, under "FAQs and Events", type a class such as ab-faq into "FAQ marker class", and give the block that holds your questions the same class. Each heading inside the block becomes a question and the text under it the answer. A definition list works too. The box starts empty, and an empty box marks nothing.
• Headings that ask. "Auto-Detect FAQ from Content" is on from the start. It reads H2, H3 and H4 headings that end in a question mark, each answered by the text under it. It skips listing pages, such as categories, tags, search results and the archive.
• YOOtheme accordions. With the YOOtheme extension switched on, an accordion whose every title is a question becomes FAQ as well, with no class typed, for as long as "Auto-Detect FAQ from Content" is on. An accordion you mark with the "FAQ accordion class", aiboost-faq unless you change it, is read whatever its titles say.

None of this brings the FAQ result back in Google, and none of it makes an assistant quote you. It writes your questions and answers in the form Bing says it uses and the assistants were shown to read.