In August 2023, Google quietly removed FAQ rich results for the vast majority of websites. Only "authoritative government and health websites" kept eligibility. For everyone else, those neat expandable question-and-answer dropdowns in the search results simply stopped appearing. Understandably, a lot of site owners and SEOs concluded that FAQ schema was no longer worth the effort. They were wrong, and this post explains exactly why.
What Google Actually Changed (and What It Didn't)
Google's announcement was specific. It removed the visual rich result for FAQ markup from organic search listings. That visual feature, the accordion of questions that expanded directly in the SERP, had been a popular click-through tool. Losing it stung.
But Google did not say FAQ schema was invalid. It did not remove FAQ from its structured data documentation. It did not penalise pages that used it. It simply stopped rendering the rich result in most cases.
That distinction matters enormously, because Google is one channel. And increasingly, it is not the only channel that reads your structured data.
How AI Search Engines Read FAQ Schema
ChatGPT, Perplexity, Gemini, and other AI-powered search tools do not render your markup visually. They parse it to understand your content. FAQ schema written in JSON-LD is machine-readable by design, which means these systems can extract question-and-answer pairs directly from your page source without needing to interpret your prose.
Think about what that means in practice. When someone asks Perplexity "how long does international shipping take from this brand?" the AI is not just scanning the text of your shipping policy page. It is also looking for structured signals that confirm and summarise the answer. A well-formed FAQ schema block on that page gives it a clean, authoritative signal: here is the question, here is the answer.
This is not speculation. Structured data has always been about communicating with machines rather than humans. The machines have simply changed. They used to be Googlebot. Now they include large language model crawlers, and those crawlers are increasingly the ones driving traffic, citations, and brand mentions that matter.
The citation logic behind FAQ schema
AI answers are built from cited sources. The models prefer pages that are easy to parse, clearly authoritative, and well-organised. FAQ schema helps on all three counts. It signals that your page has considered the questions users ask and has structured answers to them. That pattern matches exactly what an LLM is trying to do when it constructs a response: find a trusted source that answers a specific question cleanly.
A page with ten well-written FAQ schema entries is essentially handing the AI ten pre-packaged answers it can verify against your page text. That is a real advantage over a competitor whose FAQs live in a disorganised paragraph buried on a help page.
What Good FAQ Schema Actually Looks Like
Before worrying about whether schema is "dead", it is worth being clear on what good implementation looks like. A lot of the FAQ schema that was chasing rich results was low-quality: vague questions, keyword-stuffed answers, or questions that did not reflect what real users actually asked. Losing the rich result visual was, in some ways, a filter that removed the incentive to produce that kind of junk.
Solid FAQ schema for AI visibility should do the following:
- Reflect real questions. Use your support tickets, your search console queries, your Reddit threads. What do people actually ask about your product or service?
- Provide complete answers. The answer in your schema should be self-contained. Do not write "see our returns page for details" and leave it there. Give the actual answer.
- Match the page content. The schema answers should reflect what is on the page. AI crawlers cross-reference structured data against page text. Discrepancies reduce trust.
- Use plain language. Avoid jargon. AI models prefer clear, direct prose because it is easier to cite reliably.
- Limit entries to what is genuinely relevant. Eight focused, accurate FAQ entries beat twenty vague ones. Quality over quantity.
A basic JSON-LD example
Here is what a minimal, well-formed FAQ schema block looks like:
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "How long does standard delivery take?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Standard delivery takes 3 to 5 working days. Orders placed before 2pm are dispatched the same day."
}
},
{
"@type": "Question",
"name": "Can I return an item if it does not fit?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Yes. We accept returns within 30 days of delivery. Items must be unworn and in original packaging. Return postage is covered by the customer unless the item is faulty."
}
}
]
}Notice the answers are complete. They do not trail off. They give the AI everything it needs to cite your page as the source of that answer.
Where FAQ Schema Still Earns Its Place
Even setting aside AI search, there are still practical reasons to maintain FAQ schema on your site.
Voice search and featured snippets
Google's featured snippets still pull heavily from pages with clear question-and-answer structure. FAQ schema reinforces that structure in a machine-readable way. Voice search results, which pull from featured snippets, are similarly influenced. This is a smaller traffic opportunity than it was three years ago, but it is not zero.
Bing and other search engines
Bing still renders FAQ rich results. It is a smaller audience than Google, but for some industries, particularly B2B and finance, Bing's market share is meaningful. If your structured data is in place, you capture that benefit at no extra cost.
Future-proofing your content
Nobody knows exactly what Google's rich results will look like in two years. AI Overviews are already changing the SERP. The schema you maintain today may power features that do not exist yet. Removing it because one feature was deprecated is a short-sighted trade-off. The work of writing good FAQ schema is mostly done upfront. Maintaining it costs very little.
At FlinnSchema, we see this pattern regularly: brands that stripped out their FAQ schema after the 2023 announcement have been slower to appear in AI-generated answers than competitors who kept it. The structured data infrastructure built for one purpose ends up serving several others.
The Bigger Picture: Structured Data Is Now About AI, Not Just Google
The question "is FAQ schema dead?" is really a question about whether structured data matters if Google is not going to show a rich result. And the answer is that structured data was always about machine communication. Google's rich results were just one downstream application of that.
The AI search era has actually increased the value of structured data, because LLMs need reliable signals to identify authoritative, well-organised content. Schema markup is one of the clearest signals available. It is explicit, standardised, and designed exactly for the purpose of helping machines understand what a page contains.
FAQ schema sits neatly within a broader structured data strategy. If you have Offer schema on your product pages and MerchantReturnPolicy schema on your returns page, then FAQ schema on those same pages reinforces your content authority by showing that you have anticipated and answered user questions in a structured way.
It is also worth thinking about FAQ schema alongside your overall prompt research strategy. The questions you write into your FAQ schema should reflect the prompts your potential customers are entering into AI search tools. Align those two efforts and you are building a page that is specifically designed to be cited by an AI giving exactly the answer your customer is looking for.
Should You Remove Existing FAQ Schema?
No. Unless your FAQ schema is genuinely poor quality, inaccurate, or contradicts your page content, there is no reason to remove it. The risk of leaving it in place is essentially zero. The potential upside, in AI citations, in Bing rich results, in future Google features, is real.
The only scenarios where removal makes sense:
- The schema answers are outdated and you cannot keep them maintained.
- The questions are irrelevant or misleading relative to the page content.
- You have implementation errors causing validation warnings that are affecting other schema on the page.
In all other cases, keep it. Improve it if you can. Make the answers more complete and the questions more reflective of what real users ask. If you want an independent view of how your structured data is currently performing, a free AI visibility audit will show you where the gaps are.
Frequently Asked Questions
Does FAQ schema still affect Google rankings?
Not directly through rich results for most sites. Google removed the FAQ rich result visual in 2023 for non-government and non-health sites. However, well-structured FAQ content can still influence featured snippets and helps Google understand your page's relevance to specific queries. The ranking benefit is indirect but real.
Will AI search engines like ChatGPT and Perplexity read my FAQ schema?
Yes. These systems parse structured data to extract and verify information. FAQ schema in JSON-LD format is machine-readable and gives AI crawlers a clear, structured signal about the questions your page addresses and the answers it provides. Pages with well-formed FAQ schema tend to be easier for AI systems to cite accurately.
How many FAQ entries should I include on a page?
Aim for between five and ten genuinely useful questions per page. Each answer should be complete and self-contained. Avoid padding the list with vague or redundant questions just to increase the count. Quality and relevance matter far more than volume, particularly for AI citation purposes.
Should I add FAQ schema to every page on my site?
Only add it where there are genuine questions to answer. Product pages, service pages, shipping and returns pages, and help content are all natural fits. Adding FAQ schema to a homepage or a blog post where there are no real FAQs will look thin and may not provide meaningful benefit. Use it where it genuinely reflects the content and user intent of the page.