FAQ Schema for AI Search and the Patterns That Actually Get Cited
Updated: Oct 29
Key takeaways
|
FAQ schema is the AI search lever everyone has an opinion about and almost nobody implements well. Some vendors describe it as the single most powerful way to get cited in AI answers. Some technical SEOs dismiss it entirely, pointing out that language models don't read structured data the way Google does. Both camps are partly right, which is why the topic is so confusing.
I've spent a lot of time retrofitting FAQ markup across our own library and on sites I've worked with, and the honest answer sits in the middle. FAQ schema matters, but rarely for the reason people think, and almost never on its own. What moves results is treating the FAQ as a real piece of content, written for real buyer questions, with markup that simply describes it accurately.
You'll see it called FAQ schema and FAQPage schema. They're the same thing. FAQPage is the official schema.org type, and FAQ schema is what most people call it in conversation.
What is FAQ schema?
FAQ schema is structured data, written in JSON-LD using the schema.org FAQPage type, that identifies the question and answer pairs on a web page. It helps search engines understand the page's questions and answers. For AI search, its value comes from pairing valid markup with a visible FAQ section of concise answers AI systems can extract and cite. |
Why are there two pathways, not one?
The evidence looks contradictory because FAQ markup works through two different pathways. The first runs through Google. Valid FAQPage markup gives Google's crawler a structured picture of the questions a page answers, which supports how it understands and ranks the page. Since AI Overviews draw on Google's own index, that's the pathway most studies detect.
The second pathway runs through the page itself. AI assistants like ChatGPT, Claude and Perplexity read the page content, and practitioner tests suggest they treat structured data largely as text rather than parsing it the way Google does. What they respond to most is the visible content, a clear question followed by a short, direct answer that can be lifted cleanly into a response. What is an AI snippet explains why that format works so well.
Put the two together and the practical advice is straightforward. Use valid markup for Google's benefit, and make sure the same questions and answers appear clearly on the page for everyone else. Skipping either one leaves value on the table.
What about rich results?
For years, FAQ markup could earn an expanded listing in Google, with questions shown directly beneath your result. In August 2023, Google limited FAQ rich results to a small set of well-known, authoritative government and health websites. For most B2B sites, FAQ schema no longer produces that visual boost.
That doesn't make it worthless. It changes why you do it. The value now lies in clarity, helping Google understand your content and giving AI systems clean question and answer pairs to work with, rather than in a special listing. If you want a more visible win in search results, featured snippet optimization and rich snippets are better targets.
How do you write FAQPage schema correctly?
The structure is simple. A JSON-LD block declares the schema.org context and the FAQPage type. Inside it, mainEntity holds a list of Question objects. Each Question has a name, the question text, and an acceptedAnswer, an Answer object whose text holds the answer. The block goes in a script tag with type application/ld+json, like these schema markup examples.
Most of the problems I see aren't strategic. They're mechanical. A trailing comma that breaks the JSON. Quotation marks inside an answer that weren't escaped. Markup that describes questions that don't actually appear on the page, or answers that were edited on the page but never updated in the code. That last one is the most damaging, because Google's guidelines expect structured data to match visible content.
The fix is a simple habit. Write and approve the visible FAQ first, then generate the markup from that finished copy, so the two always say the same thing. Then validate it with Google's Rich Results Test and the Schema Markup Validator before publishing. Schema markup examples shows working code for FAQPage alongside Article, Organization and Person.
How do you deploy it on your site?
The principle is the same everywhere. The JSON-LD block needs to appear in the page's source code. On WordPress, Yoast and Rank Math offer FAQ blocks that generate the markup, or you can paste raw JSON-LD into a custom HTML block. On Wix, add custom structured data in the page's SEO settings. Webflow and most platforms use custom code fields, and a CMS built for AI visibility makes it easier.
Whichever platform you use, check the published page with View Page Source and confirm the markup and the visible answers are both there. If your FAQ answers only load when someone clicks to expand them, make sure the text is still in the page's HTML, because some AI crawlers don't run JavaScript. Server-side rendering for AI crawlers covers why that matters.
Which patterns get cited?
The biggest difference between FAQs that get cited and FAQs that don't is whether the questions are real. Most B2B FAQ sections reflect what marketing wished buyers would ask. The ones that earn citations come from what buyers actually ask, mined from sales calls, support tickets, People Also Ask boxes and AI assistants. AI search prompts covers how to find them.
Answers should be short and direct, usually one to three sentences that answer the question in the first line, then add one useful detail. Long, hedged answers rarely survive extraction. Each answer should stand on its own without relying on the rest of the page, the principle described in content chunking for AI. 6 to 10 questions is usually plenty for a single page. More than that tends to dilute quality. And the questions should be specific to the page, not a generic block repeated across your site, which adds nothing and can look like spam.
How do you measure whether it's working?
Because FAQ schema works indirectly, measure the outcomes rather than the markup. In Google Search Console, watch impressions and queries for pages where you've added or improved FAQs. In AI assistants, run the questions your FAQs answer and see whether your answers, or your brand, show up. Over time, track that through LLM citation tracking and your share of model.
If a page has good markup but never gets cited, look at the answers themselves before blaming the schema. In my experience, the problem is almost always that the questions aren't the ones buyers ask, or the answers don't answer them directly enough.
Where does FAQ schema fit in the Engine Optimization Matrix™?
FAQ schema sits in the Schema column of the AEO row in the Engine Optimization Matrix™, alongside Speakable and Q&A markup. Its real partner is the Content cell in the same row, question-first content with direct answers. Schema describes the content. The content does the work.
For related reading, see answer engine optimization, AI Overview optimization and how to get cited by AI.
The click is optional now. The answer isn't. The Demand Doctrine™ for AEO is the practical playbook for becoming the source behind snippets, People Also Ask and AI Overviews. Out November 18. Want all four engines? Mastering AI Visibility.
Frequently asked questions
What is FAQ schema?
FAQ schema is structured data in JSON-LD, using the schema.org FAQPage type, that identifies the question and answer pairs on a page.
Is FAQ schema the same as FAQPage schema?
Yes. FAQPage is the official schema.org type name, and FAQ schema is the common name for it.
Does FAQ schema help with AI search?
Indirectly. Google uses it to understand pages, and AI assistants respond mostly to clear visible questions and answers, so the best results come from pairing both.
Does Google still show FAQ rich results?
Only for a limited set of authoritative government and health websites, since Google restricted FAQ rich results in August 2023.
How do I write FAQPage schema?
Create a JSON-LD block with the FAQPage type and a mainEntity list of Question objects, each with a name and an acceptedAnswer containing the answer text.
How many FAQs should a page have?
6 to 10 is usually enough. Focus on real buyer questions specific to the page rather than adding more for their own sake.
What are common FAQ schema mistakes?
Invalid JSON, unescaped quotation marks, markup that doesn't match the visible content and generic FAQ blocks repeated across many pages.
How do I test FAQ schema?
Use Google's Rich Results Test and the Schema Markup Validator, then check the live page source to confirm the markup and visible answers both appear.
About the author. Harold Bell is the founder of MQL Magnet and has spent about 16 years in enterprise technology marketing, working with brands like Cisco, Rubrik and Nutanix in different capacities. He's a member of the Forbes Communications Council, the creator of the Engine Optimization Matrix™ and the author of Partner Over Product. |
