FAQ schema used to be a reliable SEO win: add the markup, get the expandable FAQ accordions in Google, claim more SERP real estate. Then Google quietly pulled the rug. Now teams chasing ChatGPT and AI visibility are asking a sharper version of the old question: does FAQ schema still work, specifically for getting cited by AI. The answer is genuinely nuanced, and getting it right saves you from both abandoning a useful practice and wasting time on a myth. Here is what actually changed, what FAQ schema does and does not do for ChatGPT, and how to use FAQ content so AI cites you.
The short answer
FAQ schema no longer earns rich results in Google for most sites, and for ChatGPT the markup itself is not what gets you cited, the visible question-and-answer content is. So if working means the old FAQ accordions in Google, no, that mostly stopped in 2023 for everyone but authoritative government and health sites. If working means helping you get cited by ChatGPT and other AI engines, the schema is at best an indirect helper, while the real driver is genuinely useful FAQ-structured content on the page that an AI can read and lift. Keep writing clear FAQ sections, because they are highly extractable. Do not expect the JSON-LD markup alone to do the work.
First, what happened to those rich results
To answer the question you have to know the history, because it shapes everything. In August 2023, Google announced changes to FAQ and HowTo rich results, restricting FAQ rich results to well-known, authoritative government and health websites. For every other site, the FAQ accordion simply stopped appearing in Search. This was not a minor tweak. Analysis afterward found that the large majority of sites with FAQ schema stopped earning FAQ rich results following the change, a dramatic drop in a visibility feature SEOs had relied on for years. Google made the change because FAQ schema had been widely abused for promotional and low-quality content rather than genuine questions. So the first thing to understand is that the classic reason people added FAQ schema, the rich result, is gone for almost everyone.
So does the schema still work?
It depends entirely on what you are asking it to do, and conflating the uses is where people go wrong.
For Google rich results: largely no, unless you are an authoritative gov or health site. The accordion is not coming back for commercial sites, so adding FAQ schema to win it is pointless.
For helping Google understand your page: marginally yes. Structured data can still help Google parse your content and reinforce entity understanding, even without producing a visible rich result.
For ChatGPT and AI citation: not directly through the markup, but yes through the content it describes. This is the crucial distinction the next section unpacks.
So FAQ schema is not dead, but its job has changed. It went from a rich-result trigger to, at most, a quiet understanding signal, while the visible FAQ content it marks up has become the thing that actually matters for AI.
How ChatGPT actually uses your content
Here is the part that reframes the whole question. AI assistants like ChatGPT read and extract the visible content of a page, the actual words a user would see, far more than they parse JSON-LD markup as structured data. Analyses of how these engines handle FAQ pages indicate that the schema markup itself is not what drives citation decisions, while the visible question-and-answer content that mirrors the schema is directly extractable by every major AI platform. In other words, when ChatGPT decides to quote your answer to a question, it is reading your on-page FAQ, not your JSON-LD. The markup is metadata that validates and describes the visible content, useful as hygiene, but it is the visible Q&A that gets lifted. This aligns with Google’s own broader guidance, which tells site owners that no special markup is required to appear in AI features and to focus on genuinely useful content.
The real value: question-and-answer content is extractable
If the markup is not the lever, why mention FAQ at all for AI. Because the FAQ format itself, a clear question followed immediately by a direct answer, is close to the ideal structure for AI extraction. AI engines quote passages that answer a specific question cleanly, and an explicit question-and-answer pair maps almost perfectly to how retrieval works. So a well-written visible FAQ section is one of the most citable formats you can publish, not because of the schema, but because of the structure. The lesson is to keep building FAQ sections, write them as genuine questions with direct answers, and treat the JSON-LD as optional supporting metadata rather than the point.
Schema versus content, side by side
The distinction is easy to blur, so here it is plainly.
| Element | Google rich results today | ChatGPT and AI citation |
|---|---|---|
| FAQPage JSON-LD markup | Rich result gone for most sites since 2023 | Not directly parsed as the citation driver |
| Visible FAQ Q&A content | Helps users; still on the page | Directly extractable and highly citable |
| Genuine, useful answers | Required for any value | The actual thing AI lifts and quotes |
| Schema not matching visible content | Violates guidelines, risks penalty | Same principle, no benefit |
Read it top to bottom and the pattern is clear: the value migrated from the markup to the visible, genuinely useful content.
What actually drives ChatGPT citations
Zoom out from FAQ specifically and the broader drivers confirm the point. Ahrefs analyzed what correlates with ChatGPT visibility and found that YouTube mentions and branded web mentions are the strongest factors, with Reddit being ChatGPT’s most-cited domain, while site page count correlates very weakly. Notice that a schema type is nowhere on that list. The things that get you cited in ChatGPT are being genuinely discussed, being fresh, and answering questions clearly, not a particular markup. FAQ-structured content helps because it answers questions clearly, which is one of those real drivers. The markup is incidental to all of it.
Should you keep the schema at all?
Yes, with the right expectations, and one firm rule. Keep FAQPage schema as low-effort hygiene where you have genuine FAQ content, because it can still help Google understand the page and costs little to maintain. But the firm rule is that the schema must match visible on-page content. Schema-only FAQs that users cannot see violate Google’s guidelines and can draw penalties, and they offer no AI benefit either since AI reads the visible content. So never add FAQ markup to a page that lacks a real, visible FAQ. Add real FAQ content first, mark it up second if you wish, and judge success by AI citations and user value, not by a rich result that is not coming back.
How to write answer content that AI cites
Make the content do the work. Use real questions, phrased the way your buyers actually ask them, including the exact searches they type into an assistant. Put a direct, two to three sentence answer immediately under each question, with the substance up front so it is liftable. Back answers with concrete specifics and, where relevant, a named source, since verifiable answers are more citable. Cover the genuine cluster of follow-up questions rather than a token three. And keep it honest and useful, not promotional, the very thing that got FAQ schema downgraded in the first place. Do that and your FAQ sections become citable assets across ChatGPT, AI Overviews, and beyond, regardless of what the markup does.
A worked example
A SaaS company had FAQ schema on dozens of pages and was frustrated that it no longer earned Google accordions and was not getting ChatGPT citations either. The schema was technically perfect. The problem was the content: short, vague, marketing-flavored answers that neither users nor an AI found useful. They rewrote the visible FAQ sections to answer real buyer questions directly, with specifics and clear language, and expanded coverage to the actual follow-up questions people ask. The Google rich result still did not return, because that feature is gone for commercial sites, but ChatGPT began citing several of the rewritten answers, because now there was genuinely extractable, useful content to lift. The same schema that did nothing for thin answers sat harmlessly atop the good ones. Content changed the outcome, markup did not.
Common misconceptions
A few myths persist. The first is that FAQ schema still earns Google rich results for everyone, when it was restricted to authoritative gov and health sites in 2023. The second is that adding FAQ schema gets you cited by ChatGPT, when AI reads the visible content, not the markup. The third is that you can skip the visible FAQ if you have the schema, when schema must match visible content and AI ignores invisible markup anyway. The fourth is that FAQ content is therefore useless, when in fact FAQ-structured visible content is one of the most extractable, citable formats for AI. Hold the accurate version: the markup faded, the well-written visible FAQ became more valuable.
How to find the questions worth answering
Since the value is now in answering real questions well, start by knowing which questions matter. Map the genuine queries your buyers ask assistants in your category, then write FAQ content that answers them directly. I use SQSEO for this, fanning one seed keyword into the longtail and question-level prompts that trigger AI answers, for free, so your FAQ sections answer prompts people actually type rather than ones you guessed. The related question of whether any schema is required for AI Overviews is covered in do you need specific schema markup for AI Overviews, and the factor-level view of getting cited is in Google AI Overview ranking factors.
The bottom line
Does FAQ schema still work for ChatGPT SEO? The markup does not get you cited, and it no longer earns Google rich results for commercial sites, so if that was the plan, no. But FAQ-structured visible content works better than ever, because the explicit question-and-answer format is exactly what AI engines extract and quote. Keep your FAQ sections, write them as genuine questions with direct, useful answers, add the schema as optional hygiene that matches the visible content, and measure success in AI citations rather than a rich result that retired in 2023. The format endures, the markup-as-magic never existed.
Frequently asked questions
Does FAQ schema still work for ChatGPT SEO?
Not the markup itself. ChatGPT reads and extracts your visible question-and-answer content, not your FAQPage JSON-LD, so the schema is at most an indirect helper. What works is genuinely useful, visible FAQ content written as clear questions with direct answers, which is one of the most extractable, citable formats for AI. Keep the FAQ content, treat the markup as optional hygiene.
Do FAQ rich results still appear in Google?
For most sites, no. In August 2023 Google restricted FAQ rich results to well-known, authoritative government and health websites, and the vast majority of other sites stopped earning the FAQ accordion. So adding FAQ schema to win a Google rich result is no longer effective for commercial sites.
Should I remove my FAQ schema?
No need, if it matches visible content. It is low-effort hygiene that can still help Google understand your page. The firm rule is that the schema must match a real, visible FAQ on the page, since schema-only FAQs violate guidelines and give no AI benefit. Add visible FAQ content first, mark it up second.
What actually gets my FAQ answers cited by AI?
Clear, genuine answers to the real questions people ask, placed directly under each question so they are easy to lift, backed by specifics, and covering the true cluster of follow-up questions. Being discussed across the web and staying fresh help too. The format and the usefulness drive citations, not the markup.
