Schema and FAQ content matter for AI search. For B2B companies in manufacturing, SaaS, and industrial services, schema markup for AI search is not a trick that forces a ChatGPT citation or an AI Overview. It is part of a system: the questions you model from real buyers, the answers you make visible on the page, the entity you prove, and the structured data you keep accurate and fresh. That system is the OneIMS MAPS framework: Model Buyer Intent, Answer Clearly, Prove & Place, and Structure & Stay Fresh.
Four stages. Each one gives the markup something real to describe.
Model Buyer Intent
FAQ schema only matters if the questions are the questions your buyers ask. Read the Model Buyer Intent stage.
Answer Clearly
Visible, self-contained FAQ answers are what AI systems can extract. Read the Answer Clearly stage.
Prove & Place
Entity schema makes it unambiguous who is speaking and what is being offered. Read the Prove and Place stage.
Structure & Stay Fresh
After May 7, 2026, FAQ markup serves understanding, not Google SERP accordions. Read the Structure and Stay Fresh stage.
Google fully removed FAQ rich results on May 7, 2026, as recorded in Google's Search Central documentation updates and reported by Search Engine Land. The accordion display is gone. The work is not. FAQ schema now supports understanding, not a SERP decoration, and it only does that job when it describes visible answers to questions your buyers actually ask.
This article shows how each MAPS stage uses FAQ content and structured data for AI, and the proof behind it. The honest frame is this: Google says structured data is not required for AI Overviews or AI Mode, while Microsoft has said schema helps its LLMs understand content, and the markup is there to reinforce the visible answers, not to replace them.
Key takeaway: Schema and FAQ content matter for AI search when they sit inside MAPS. They make your questions, answers, and entity easier to understand. They are not a shortcut around clear content.
Short answer: Schema markup for AI search is structured data, usually JSON-LD in the schema.org vocabulary, that tells machines what a page means: which company is speaking, what service or product is being described, who wrote it, and which visible questions and answers it contains. FAQ schema is one part of that layer. It matters because it reinforces clear, on-page answers. It does not replace those answers, and it is not a special ticket into Google AI Overviews.
HTML tells a browser how to render a page. Schema states the same facts in a consistent vocabulary: this is an organization, this person is the author, this is the service, this is the accepted answer to this question. Without that layer, a system has to infer more of that meaning from prose, which leaves more room for error. With it, the visible content and the machine-readable description can agree.
That is also why schema only works as part of MAPS. If the questions are the wrong questions, if the answers are buried or missing from the page, or if the entity is inconsistent, the markup has nothing reliable to describe. The OneIMS AEO guide covers that broader practice. Here, the focus is how FAQ content and structured data prove their value at each MAPS stage.
Short answer: Google has not said that FAQ schema creates or improves AI Overviews. In its guide to optimizing for generative AI features, Google writes: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add. However, it's a good idea to continue using it as part of your overall SEO strategy, as it helps with being eligible for rich results on Google Search." Keep FAQ schema because it documents visible answers and supports a clearer page. Do not keep it because you expect it to unlock AI Overviews.
Google's AI features and your website guidance adds that there are no additional requirements to appear in AI Overviews or AI Mode, that you do not need special machine-readable files or markup, and that any structured data you use should match the visible text on the page.
Microsoft's public position is different in emphasis, not a contradiction of that Google point. At SMX Munich on March 18, 2025, Fabrice Canel, Principal Product Manager at Microsoft Bing, said that schema markup helps Microsoft's LLMs understand content, as reported by Search Engine Land. That is a reason to implement accurate schema. It is not evidence that Google uses the same markup as an input to AI Overviews.
The comparison below is the shortest way to separate the system from the shortcut:
Each card carries a Claim or tactic, then What the evidence supports.
Adding FAQ schema will make an accordion appear in Google
No. Google removed FAQ rich results for all sites on May 7, 2026, including the government and health sites that had remained eligible since August 2023.
FAQ schema is therefore useless
No. FAQPage remains a valid schema.org type. Google has said unused structured data does not cause problems for Search and does not need to be removed. The markup can still describe visible questions and answers.
Hidden or JSON-LD-only answers are enough, because AI will parse the schema
No. An independent test first published in December 2025 found that the systems tested used visible HTML. None named a fact that existed only in JSON-LD.
Schema is a special requirement for AI Overviews
No. Google says structured data is not required for generative AI search, and that there is no special schema.org markup to add.
Schema and FAQ content do not matter at all
No. Google still uses structured data for remaining rich results and requires it to match visible content. Microsoft has said schema helps its LLMs understand content. Visible, self-contained FAQ answers are what live-fetch systems have been shown to extract.
Schema plus visible FAQ answers inside MAPS
This is the supported stance. The questions have to be real buyer questions, the answers have to be on the page, the entity has to be clear, and the markup has to stay accurate and fresh.
HowTo rich results follow the same pattern. Google limited them to desktop in August 2023 and fully deprecated them as of September 13, 2023, in the same Search Central Blog post that first restricted FAQ rich results. The HowTo type remains valid in schema.org. Like FAQ schema, it now serves structure and understanding, not a Google rich result.
Both rich results followed the same path, from available, to limited, to gone. The vertical axis is the status Google documented at each date.
| Date | FAQ rich results | HowTo rich results |
|---|---|---|
| Before August 2023 | Available | Available |
| August 2023 | Limited to government and health sites | Limited to desktop |
| September 13, 2023 | Limited to government and health sites | Removed |
| May 7, 2026 | Removed | Removed |
Short answer: FAQ schema only matters if the questions are the questions your buyers ask. Model Buyer Intent means those questions come from sales conversations, support tickets, and the prompts buyers type into AI tools, not from a generic FAQ template. The markup then labels a real question-and-answer set. It cannot invent buyer intent that is missing from the page.
B2B research rarely starts with your product name. A manufacturing operations leader asks what an industrial automation integrator actually does. A SaaS buyer asks how to compare two implementation approaches. An industrial services director asks what is included, who it is for, and how delivery works. Those are the questions an assistant will be asked, and they are the questions your FAQ should answer.
Build the question set from sources you already have:
Then group them by stage: awareness, problem-solving, comparison, and selection. A page that answers "what does this service include?" with FAQPage markup is doing MAPS work. A page that marks up "Why is [Company] the best?" is not. That is marketing copy wearing a question mark, and it is a misuse of FAQ schema.
The proof at this stage is practical, not theatrical. Schema can only describe what you chose to ask and answer. If the questions do not match buyer and AI prompts, you will have technically valid markup on commercially irrelevant content. If they do match, the rest of MAPS has something worth making clear, visible, and machine-readable.
Short answer: Visible, self-contained FAQ answers are what AI systems can extract. An independent test first published in December 2025 found that the systems tested used facts from visible HTML. None named a fact that existed only in JSON-LD. FAQ schema matters when it matches those on-page answers, not when it tries to carry facts the page never shows.
That is the proof for Answer Clearly, and it should change how B2B teams write FAQs. Every question and every answer in your FAQPage block must appear on the page itself. Google already requires structured data to match visible content. The live-fetch evidence says the same thing from the other direction: if a qualifying detail, a service scope, or a definition lives only in the JSON-LD, a system that fetches the page may not see it.
Source: independent SEO test, Dec 2025.
Write answers that can stand alone:
Aim for roughly 40 to 60 words. That is practical writing guidance, not a finding from a published study. Long answers are harder to quote cleanly. Very short ones often lack the specifics a buyer, or an assistant summarizing your page, needs.
The same December 2025 test has limits you should keep in view. It was mainly a direct-fetch experiment on a single page, with five systems queried on October 30, 2025: ChatGPT, Claude, Gemini, Perplexity, and Google AI Mode. It has not been published as a repeated test, and it does not answer whether Google AI Overviews or Bing Copilot use schema from an index. It is not proof that "AI systems never use schema." It is proof that visible answers matter, and that FAQ schema without visible answers is a hollow layer.
This is why schema and FAQ content matter together. The on-page answer is what a live-fetch system can read. The schema is the consistent label for that same answer. Answer Clearly is the stage that makes both of them useful.
Short answer: Organization, Person, Product, and Service schema make it unambiguous who is speaking and what is being offered. Microsoft Bing's Fabrice Canel said at SMX Munich in March 2025 that schema markup helps Microsoft's LLMs understand content. Google still uses structured data for remaining rich results and requires that markup to match visible content. Prove & Place is where those entity signals, and the off-page proof they point to, earn their keep.
When a buyer asks an assistant for a type of B2B provider, the system has to decide which company you are, what you actually offer, and whether other sources corroborate that. Schema does not create that reputation. It does make the on-page identity harder to misread.
The types that do this work for a B2B site are:
| Schema type | Where it belongs | How it proves the entity |
|---|---|---|
Organization with sameAs | Homepage, used consistently sitewide | Names the company and links it to verified profiles such as LinkedIn, a Google Business Profile, or Crunchbase, so the website entity can be matched to the same company elsewhere |
Person / author markup | Every article and resource | Attributes the content to a named, credentialed person, which supports the Experience, Expertise, Authoritativeness, and Trustworthiness (E-E-A-T) framework in Google's Search Quality Rater Guidelines |
Article / BlogPosting | Editorial content | Supplies headline, author, publisher, and publish or modified dates so claims can be attributed and dated |
Service | Solution and service pages | States what you offer, who it is for, and who provides it. This is the priority type for industrial, professional, and other B2B service companies |
Product | Product or packaged-offer pages | States the product name, description, and, where relevant, offer details. It can still support product rich results in Google when guidelines are met |
FAQPage | Pages with genuine, visible Q&A | Ties specific questions to accepted answers from the same organization, as long as those answers are on the page |
BreadcrumbList | Inner pages | Makes the site hierarchy explicit and can still produce breadcrumb rich results in Google |
Nothing ticked yet. Start with one consistent Organization entity and author markup on every article.
Do not ship two conflicting Organization blocks, one from the CMS and one added by hand. Use one consistent entity, and populate sameAs with the profiles you actually control. Then do the placement work that schema cannot do alone: industry directories, trade publications, partner pages, and other third-party mentions that repeat the same company name and offer. The markup points to those placements. The placements are the proof.
The verified record for this stage is straightforward. Google continues to use structured data to make pages eligible for rich results, and it requires that structured data to match the visible page. Microsoft, through Canel at SMX Munich, has said schema helps its LLMs understand content, and the same session emphasized that generative AI systems value fresh content. Neither statement is a promise that schema will get you named in ChatGPT or in an AI Overview. Together, they are a strong reason to treat entity schema as part of the system, not as optional decoration.
Short answer: After May 7, 2026, FAQ markup serves understanding, not Google SERP accordions. Structure & Stay Fresh means consistent JSON-LD that matches the visible page, honest dateModified values, and regular validation. QAPage is still a supported Google type for user-submitted Q&A. FAQPage remains valid schema.org markup for publisher-written FAQs, even though it no longer produces a Google rich result.
That May 7, 2026 change is the structural fact every B2B team needs in view. Google recorded the deprecation in its documentation updates on May 8, 2026, and removed the FAQ rich-result documentation on June 15, 2026. A notice on that documentation, quoted by Search Engine Land, also said Search Console reporting and Rich Results Test support for FAQ would be dropped in June 2026, and Search Console API support in August 2026. Your markup is not broken. The display feature is gone.
What to do with that fact:
FAQPage block.FAQPage. QAPage is for pages where users submit answers to a question. Google still documents QAPage, and on March 24, 2026, it added more supported properties for Discussion Forum and QA Page markup.Article or Service, for example), with FAQ markup only where a visible FAQ exists. The JSON-LD and the HTML must say the same thing.dateModified when you actually revise the content. Canel's SMX Munich remarks also noted that generative AI systems value fresh content, and that IndexNow can notify Bing when pages are published or updated.FAQPage and HowTo. Use Google's Rich Results Test only for types that can still produce Google rich results, such as Product or Breadcrumbs. The Rich Results Test is no longer the right tool for FAQ markup.After you update a page, request indexing in Google Search Console and monitor the remaining rich-result reports for errors. Recrawling can take time. A clean report means the markup is being read without issues. It is not, by itself, a measure of AI citations.
To see whether the system is working in the answers that matter, track a fixed set of buyer-intent prompts and log brand mentions and citations. That measurement is the subject of our guide to AI share of voice. Schema validation and mention tracking answer different questions. You need both.
Short answer: Use this self-audit to score your own site against the four MAPS stages. It is a checklist, not a published benchmark. Give your site 1 point for each item you can honestly mark yes, for a maximum of 12. The same items can power a scored quiz in an interactive version of this article.
| Stage | Score |
|---|---|
| Model Buyer Intent | 0 of 3 |
| Answer Clearly | 0 of 3 |
| Prove and Place | 0 of 3 |
| Structure and Stay Fresh | 0 of 3 |
Tick the items above to score your site. Treat the number as a diagnostic for your team, not as an industry ranking.
Get help closing the gapsHow to read your score: 10 to 12 means the system is in place and the work is mostly maintenance. 7 to 9 means the foundation is there, but one stage is weak, often buyer-intent coverage or off-page proof. 0 to 6 means schema may exist, but it is not yet doing MAPS work. Treat the number as a diagnostic for your team, not as an industry ranking.
These eight answers are visible on the page, and the markup describes those same answers.
Select "See the markup" to build this from the FAQ below.
This sample is generated from the questions and answers you can read below it, which is the whole point: the markup describes visible content rather than carrying facts the page never shows. Answer text is shortened here so the block stays readable.
Yes. Schema and FAQ content matter for AI search as part of the MAPS system, not as a shortcut. Visible, self-contained answers are what an independent December 2025 test found AI systems could extract. Microsoft has said schema helps its LLMs understand content. Google still uses structured data for rich results and requires it to match visible content. Google has also said structured data is not required for generative AI search, so the markup reinforces the page. It does not replace it.
Google has not said that FAQ schema helps you appear in AI Overviews. Its official guidance is that structured data is not required for generative AI search and that there is no special schema.org markup to add. FAQ schema still matters when it matches visible answers on the page, and Google still recommends structured data for rich-result eligibility. Do not implement FAQ schema in order to "optimize for AI Overviews." Implement it to describe real Q&A content clearly.
Use MAPS. Model Buyer Intent by writing FAQ questions from sales, support, and real AI prompts. Answer Clearly with visible, self-contained answers. Prove & Place your entity with consistent Organization, author, and Service or Product schema, plus off-page corroboration. Structure & Stay Fresh with valid JSON-LD, honest dateModified values, and regular validation. The OneIMS AEO guide walks through that practice in more depth, and AEO and GEO services are how we implement it for B2B companies.
OpenAI has not published documentation that ChatGPT's crawlers parse JSON-LD, and there is no evidence that adding a schema type directly causes ChatGPT to cite your page. The practical path is still MAPS: answer real buyer questions in visible text, make the organization and authors unambiguous, place corroborating proof off-site, and keep the content fresh. ChatGPT Search has been reported to draw in part on Bing's index, and Fabrice Canel of Microsoft Bing said at SMX Munich in March 2025 that schema helps Microsoft's LLMs understand content. That supports accurate, visible markup. It does not guarantee a mention.
Lead with service-page clarity, then mark up the same visible content. On each solution page, state what the service is, who it is for, and how it is delivered. Add Service schema and a consistent Organization entity with sameAs links, and include a genuine FAQ for the questions buyers actually ask about that service. Pair the on-site work with placements in industry directories, trade publications, and partner pages. Schema describes the service. Clear answers and off-page proof make the company easier to recognize and recommend.
Yes, with the right expectations. Google fully removed FAQ rich results on May 7, 2026, so no B2B site will see an accordion in Google Search. FAQ schema still matters because FAQPage remains valid, Google has said unused structured data does not cause problems for Search, and visible question-answer content is what live-fetch systems have been shown to use. Keep the markup if the FAQ is genuine and on the page. Do not add it to game a display feature that no longer exists.
No. Google's 2023 guidance states that unused structured data does not cause problems for Search and does not need to be removed proactively. If your FAQ content is genuine and visible, the markup accurately describes the page and can stay. There is no penalty for keeping valid, matching FAQ schema. Remove it only if it is inaccurate, hidden, or being used to mark up marketing copy.
Start with the MAPS scorecard in this article and fix any stage that scores low. Then validate the markup with the Schema Markup Validator, and use Google's Rich Results Test only for types that can still produce rich results. After deployment, monitor Search Console's remaining rich-result reports for errors. To judge AI visibility, run a fixed set of buyer-intent prompts and track mentions and citations, as described in our AI share of voice guide. A clean schema report and a rising mention rate are related, but they are not the same metric.
Schema and FAQ content matter for AI search. They matter more when they are treated as a system.
Model the questions your buyers already ask. Answer them in visible, self-contained language. Prove the entity with consistent organization, author, and service or product markup, and place that same entity where other sources can corroborate it. Structure the JSON-LD so it matches the page, and keep it fresh. That is MAPS. That is why the markup is still worth doing after the Google FAQ accordion disappeared.
If you want to see where your B2B site stands across all four stages, OneIMS AEO and GEO services are built for manufacturing, SaaS, industrial, and professional services companies. You can also start with a free baseline of your AI search readiness through OneIMS GrowthScore.
See where your site stands across all four MAPS stages.
Start with a structured review of buyer-intent coverage, answer clarity, entity proof, and structured data. Or get a free baseline of your AI search readiness first.