Most FAQ pages are written for the owner, hidden inside an accordion, or marked up with answers that no longer match the page, and an AI engine quotes none of them. An FAQ an AI can read is plain visible text: a question in the customer's words as a heading, a complete answer under it, and the same words in a FAQPage script if you add one. Google retired FAQ rich results on 7 May 2026 and says structured data is not required for its AI features, so the visible text is what counts. Write the questions, answer each one so it stands alone, put it on the page, paste the markup, validate it, and keep the two copies identical.
A customer asks ChatGPT whether a clinic in Lekki opens on Saturdays, or whether a Surulere tailor can finish an agbada in a week, or whether a Lagos freight agent handles customs clearance at Apapa. The engine looks for a page that answers the question in one place. An FAQ page written in the customer's words is the page most likely to hold that sentence, and most FAQ pages fail at it: the questions are the owner's, the answers hide inside an accordion, or the markup says one thing while the page says another.
This post is for whoever edits the website. It gives the six steps, the block of markup to paste, and the one rule that keeps the markup honest. It also says plainly what Google does with FAQ markup in 2026, because that changed this year and most guides have not caught up.
What does "an FAQ that AI can read" mean?
It means visible text, in a fixed shape: a question a customer would type, as a heading, and a complete answer in the paragraph under it, readable without clicking anything. The markup, a FAQPage script in the page's code, is a machine-readable copy of that text. It is useful and it is optional. The text is the thing.
Google's guide to its generative AI features, last updated 10 July 2026, says that structured data is not required for generative AI search and that there is no special schema.org markup a site needs to add. OpenAI's crawler documentation, read on 26 October 2026, does not mention structured data at all, and neither does Perplexity's. So no engine documents reading FAQ markup to build an AI answer. Every engine reads the page. Write for the page first.
Does Google still show FAQ rich results?
No. Google's own documentation-updates page carries two entries. On 8 May 2026, under "Deprecating the FAQ rich result feature", it says the feature "will no longer appear in Google Search starting May 7, 2026". On 15 June 2026 it records that the documentation for the feature was removed. As of 26 October 2026, Google's gallery of structured data features lists twenty-five types, and FAQ is not one of them.
The dropdown had been shrinking for years. The same updates page records, on 14 September 2023, that the FAQ rich result was limited to "well-known, authoritative government and health websites". A Lagos business never had it after that date. So nothing a small business could see has been lost, and the reason to keep an FAQ page was never the dropdown. It was, and is, the answer.
Why add the markup at all, then?
Two reasons, both modest. First, schema.org is a vocabulary every engine can parse, and Google's introduction to structured data describes it as a standardised format for providing information about a page and classifying its content. A FAQPage block tells any machine, in one place, that these sentences are questions and these are their answers. Second, writing the block forces the discipline that matters: each answer has to be a standalone piece of text that makes sense out of context, which is exactly what an AI answer needs to quote.
What the markup does not do is substitute for the visible text, and Google is specific about that. Its structured data guidelines, last updated 10 July 2026, say: "Don't mark up content that is not visible to readers of the page", and "Your structured data must be a true representation of the page content". A page that breaks those rules can lose eligibility for rich results through a manual action. Keep that in mind at step four.
Step 1: Collect the questions customers actually ask
Open WhatsApp and your call log. Write down the questions you have answered more than three times this month, in the words the customer used. "Do you open on Saturdays?" "How long does an agbada take?" "Do you deliver to Ajah?" "Can I pay on delivery?" Five to ten of these is the right number for one page; a question nobody asks is filler, and filler makes the real answers harder to find.
Leave out questions whose answer is "contact us". The engine cannot quote a redirect.
When this step is done, you will have a list of questions in the customer's words, with the one most people ask at the top.
Step 2: Write each answer so it stands alone
An AI answer may quote one paragraph with nothing around it. So each answer has to carry its own context: the business name, the place, the number, the condition. Two to four sentences. The first sentence answers the question outright; the rest add the specifics a machine can repeat.
A made-up example, labelled as such so nobody goes looking for it. Question: "Do you open on Saturdays?" Weak answer: "Yes, we do!" Strong answer: "Yes. Example Dental in Lekki Phase 1 opens 9am to 2pm on Saturdays for check-ups, cleaning and emergencies. Book by WhatsApp before Friday evening for a Saturday slot." The second one can be read by itself and still be correct.
The research points the same way. In a paper presented at KDD 2024, Aggarwal and colleagues tested which changes to a page made generative engines more likely to use it, and the authors report that adding citations, quotations from sources, and statistics raised a page's visibility in the answers by up to 40% in their benchmark. Concrete, attributable sentences get used. "Yes, we do!" does not.
When this step is done, every answer can be pasted into a chat on its own and a stranger could act on it.
Step 3: Put the questions and answers on the page as plain text
Each question is a heading, each answer is the paragraph under it, and the whole thing is visible when the page loads. That is the shape every engine can read. Google's AI guide says there is no requirement to break content into tiny pieces for AI; a heading and a paragraph is plenty.
If the site uses an accordion that opens on click, two conditions apply. The answer text must be in the page's HTML, so a crawler that does not run scripts still sees it, and the accordion must actually work. We learned the second one on our own site in August 2026. The stylesheet collapsed every FAQ answer to zero height, the script that was supposed to open them had never been added, and the FAQPage markup, added in July, claimed six answers no visitor could read. That is the exact case Google's guidelines describe. The fix was a few lines of script and a rule that the FAQ is open by default, and the safest FAQ is one that never closes.
When this step is done, you can view the page source in any browser, search for a full sentence from one of your answers, and find it.
Step 4: Paste the markup
This is the block. Copy it, replace the two example questions and answers with your own, and keep adding items inside the square brackets for each further question. It goes anywhere in the page's code; Google's documentation says JSON-LD can sit in a script tag in the head or the body.
<script type="application/ld+json">{"@context":"https://schema.org","@type":"FAQPage","mainEntity":[{"@type":"Question","name":"Do you open on Saturdays?","acceptedAnswer":{"@type":"Answer","text":"Yes. Example Dental in Lekki Phase 1 opens 9am to 2pm on Saturdays for check-ups, cleaning and emergencies. Book by WhatsApp before Friday evening for a Saturday slot."}},{"@type":"Question","name":"Do you accept HMO cards?","acceptedAnswer":{"@type":"Answer","text":"Yes. Example Dental accepts the HMO plans listed on our pricing page. Bring the card and a form of ID to the first visit."}}]}</script>
Four parts, and nothing else is needed: the type FAQPage, a list called mainEntity, one Question per item with the question in "name", and the answer inside acceptedAnswer as "text". Schema.org defines FAQPage as "a WebPage presenting one or more Frequently asked questions", and that is the whole vocabulary you are using.
Where to paste it without a developer. On WordPress, add a Custom HTML block anywhere on the page and paste the whole thing in. On Wix, Squarespace, Webflow and most builders, each page has a setting for custom code, usually called code injection or header code, and it goes there. On a hand-coded site, it goes before the closing head or body tag.
The one rule: the text in the script must be identical to the text on the page, question for question and word for word. Not a shorter version, not a tidier one. Identical.
When this step is done, the page source contains your script and your questions, and the page looks exactly the same as before.
Step 5: Validate it
Two free checks, in this order. First, the Schema Markup Validator at validator.schema.org: paste the page's address, and it parses every script block on the page and lists any error in the vocabulary, usually a missing bracket or a stray comma from editing by hand. Second, Google's Rich Results Test, which Google's own documentation recommends for checking that it can read the page's structured data. Since the FAQ rich result has been retired, do not expect that tool to report an FAQ item; use it to confirm the page parses and nothing else on the page is broken.
When this step is done, the validator shows zero errors on your page.
Step 6: Keep the two copies in step
The markup goes wrong months later, when someone changes the Saturday hours on the page and forgets the script, and now the machine-readable copy says 2pm while the visible copy says 4pm. Google's guideline about structured data being a true representation of the page is the one that bites here. Add one line to whatever checklist the person who edits the site uses: if you change an FAQ answer, change it in both places. Once a month, run the validator again and read the two copies side by side.
When this step is done, there is one person who knows the rule and a date in the calendar.
What an FAQ gets you, and what it does not
It gets you the page most likely to hold the exact sentence a customer's question needs, in a shape every engine reads. It does not get you a place in any answer; nobody can promise that. Google's AI features page says a page has to be indexed and eligible for a snippet to appear in AI Overviews or AI Mode, with no additional technical requirements, and OpenAI publishes no rule at all beyond letting its crawler in. The check is the same as for every other page: ask the engines the five questions on the list, on the same day each month, and write down who gets named.
The demand is there to be answered:
Those customers ask a question and read one answer. The FAQ page is where you write it.
Want to know what the engines already say about your business?
Our free AI Visibility checklist is the 12-point check we run before we quote, and the FAQ page is one of the twelve. It starts with what ChatGPT, Gemini and Google say when a customer asks about a business like yours. Enter your email and we will send it over.
✓ Check your inbox, your checklist is on the way!
We respect your privacy. Unsubscribe at any time.
The FAQ an AI can read is the one a customer can read without clicking, and the markup is a copy of it, never a substitute. Google no longer shows FAQ rich results and does not require the markup for its AI features; the markup still costs ten minutes and still has to match the page word for word.
One forward before you go: open your own website's FAQ, view the page source, and search for one full sentence from an answer. If it is not there, no machine can read it either, and that is the first thing to fix before any markup goes in.
Frequently asked questions
Is FAQ schema dead now that Google retired the rich result?
The rich result is gone; the vocabulary is not. Google's documentation-updates page says the FAQ rich result feature stopped appearing in Google Search on 7 May 2026, and the documentation for it was removed on 15 June 2026. FAQPage is still a valid schema.org type that any engine can parse. What died is the dropdown under a Google result, which had been limited to government and health sites since September 2023 anyway.
Does ChatGPT read FAQ schema?
OpenAI's crawler documentation, read on 26 October 2026, says nothing about structured data of any kind. Neither does Perplexity's. Google's guide to its generative AI features says structured data is not required for generative AI search. So no engine documents reading FAQPage markup for its AI answers. What every engine does read is the visible text on the page, which is why the text comes first in this post and the markup comes fourth.
Do I need a plugin or a developer to add the markup?
No. The markup is one script block of text. On WordPress it goes into a Custom HTML block on the page; on Wix, Squarespace and most site builders there is a setting for custom code on a page or in the site header. Paste the block, replace the example questions and answers with your own, save, and check the page source. A plugin that generates it from your visible text is fine too, as long as you check that its output matches what the visitor sees.
Can I keep the FAQ in an accordion that opens on click?
Only if the answer text is in the HTML and readable without scripts, and only if the accordion works. Google's structured data policies say not to mark up content that is not visible to readers of the page. Our own site broke this rule in August 2026: a stylesheet collapsed every answer to zero height, the toggle script had never been added, and the markup claimed six answers no human could see. The safest FAQ is open by default.
How many questions should an FAQ have?
As many as customers actually ask, which is usually five to ten. Each one has to be a real question with a complete answer, and the page has to be about more than one question. A page padded with questions nobody asks is harder to quote, because the engine has to find the one sentence that matters inside the filler.
Is this the same as SEO?
It overlaps. A visible FAQ in the customer's words has been good SEO for years, and Google's own AI guide says the work for its AI features is still SEO. The difference is what you measure. SEO measures rankings and clicks; AI visibility measures whether the answer to the customer's question names your business. The FAQ page serves both, and the second one has to be checked by asking the engines directly.
Will an FAQ get my business into Google's AI Overviews?
Nobody can promise that. Google's AI features page says a page has to be indexed and eligible for a snippet to appear in AI Overviews or AI Mode, with no additional technical requirements. A clear FAQ makes a page easier to quote; it does not buy a place. Google's Generative AI performance report in Search Console shows whether any of your pages appeared, and that is the number to watch after the FAQ goes live.
References & sources
- Google Search Central, Search Central documentation updates (read 26 October 2026. Entries of 8 May 2026 (deprecating the FAQ rich result feature; 'will no longer appear in Google Search starting May 7, 2026'), 15 June 2026 (documentation removed) and 14 September 2023 (feature limited to government and health websites))
- Google Search Central, Structured data features in Google Search (the search gallery) (read 26 October 2026; last updated 16 October 2026. Twenty-five features listed; FAQ is not among them)
- Google Search Central, General structured data guidelines (last updated 10 July 2026. 'Don't mark up content that is not visible to readers of the page'; 'Your structured data must be a true representation of the page content'; manual action)
- Google Search Central, Introduction to structured data markup in Google Search (last updated 10 December 2025. JSON-LD recommended; script tag in head or body; the Rich Results Test)
- Google Search Central, Guide to optimizing for generative AI features on Google Search (last updated 10 July 2026. 'Structured data isn't required for generative AI search'; 'There's no requirement to break your content into tiny pieces')
- Google Search Central, AI features and your website (eligibility: indexed and eligible for a snippet, no additional technical requirements)
- Schema.org, FAQPage ('A FAQPage is a WebPage presenting one or more Frequently asked questions')
- Schema.org, Schema Markup Validator
- Google, Rich Results Test
- OpenAI, OpenAI bots documentation (OAI-SearchBot, OAI-AdsBot, GPTBot, ChatGPT-User) (read 26 October 2026; says nothing about structured data)
- Perplexity, Perplexity crawlers (PerplexityBot, Perplexity-User) (read 26 October 2026; says nothing about structured data)
- Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- Pew Research Center, Americans and AI 2026 (5,119 US adults, February 2026)
- BrightLocal, Local Consumer Review Survey 2026 (1,002 US adults)