Google restricted FAQ rich results to a narrow set of government and health sites in August 2023. For nearly every other website, the FAQ accordion disappeared from the search results page overnight. Marketers who had spent years building FAQ sections for that visual space asked the obvious question: is FAQ schema dead? It is not. The reader changed. You are no longer writing an FAQ answer for a Google rich result. You are writing it for ChatGPT, Perplexity, and Google AI Overviews to lift directly into a generated response.
I hear a version of this question from nearly every client who remembers the old FAQ snippet. They see the accordion has vanished from their search listing and assume the underlying work stopped mattering. It did not stop mattering. It changed audience. AI models read a well-built FAQ section the same way a person skimming for a fast answer would: a clear question, followed by a complete answer, with nothing else required to make sense of it.
This guide covers what changed with FAQ rich results, why FAQPage schema still belongs on your site, and exactly how to write FAQ content that AI engines can lift word for word. You will get the structure, the schema, the placement rules, and the mistakes that quietly keep well-built FAQ sections from ever getting cited.
The distinction matters more than most content teams realize. Buyers who used to type a query into Google now open ChatGPT or Perplexity and ask the question directly. If you stripped FAQ sections out of your content strategy because the rich result disappeared, you removed the exact format most likely to earn a citation from the tool your buyers now consult first.
Why FAQ rich results disappeared from Google Search
For years, FAQPage schema was one of the highest-value additions in SEO. Mark up a handful of questions and answers, and Google would often render them as an expandable accordion directly inside the search results, sometimes doubling the space a single listing took on the page. Click-through rates went up. Every site with an SEO budget added an FAQ section, whether or not the content answered a real question anyone had asked.
In August 2023, Google announced it would limit FAQ rich results to a small set of authoritative government and health websites. For the rest of the web, the accordion stopped appearing. Google did not ban FAQPage schema. It stopped rendering the schema as a rich result for most sites, because the format had filled up with FAQ sections built purely to win the space rather than to answer anything useful.
FAQ schema still matters, it just answers to a different reader
The FAQ rich result was built for one reader: a person scanning a page of search results. FAQPage schema today is read by a second, much larger audience: the AI models generating answers inside ChatGPT, Perplexity, Google AI Overviews, and Copilot. That audience never went away. If anything, it grew faster than the rich result ever did.
Schema markup does not force an AI model to cite you, but a well-structured FAQ section is consistently one of the fastest ways to earn a citation, because it hands the model a complete answer with no parsing required. We cover the full case for structured data in our guide to schema markup for AEO, but the short version for FAQ content specifically: FAQPage schema provides pre-structured question and answer pairs, and pre-structured answers are exactly what an AI model is trying to extract when it builds a response.
The rich result is gone. The reader is not. FAQ content and FAQPage schema now work for AI answer engines instead of the Google results page, and the rules for writing to that reader are different from the rules that won the old accordion.
How AI engines actually read your FAQ section
AI models do not render your schema visually the way Google once did. Most read the page content directly and treat FAQPage schema as a confirming signal rather than the primary source. That distinction changes how closely your schema and your visible text need to match.
Under the old rich-result rules, the schema text and the visible page text could drift apart without much consequence, since Google displayed whatever sat inside the schema regardless of what a visitor saw on the page. Under AI citation rules, the two need to match closely. If your FAQPage schema states one answer and your visible page says something slightly different, you have handed the model conflicting signals. The mismatch itself makes your content harder to trust and cite, whichever version the model happens to read.
Confirm AI crawlers can see the FAQ content at all before you worry about anything else. A section built with client-side JavaScript that never renders in the initial HTML is invisible to a crawler that does not execute scripts. A perfectly worded FAQ section that never loads for GPTBot or ClaudeBot cannot get cited, no matter how well it answers the question.
Different engines weight this content differently too. Perplexity leans heavily on freshly crawled pages and often pulls FAQ answers close to verbatim, with a link back to the source. ChatGPT blends training data with live retrieval, so an established FAQ answer can surface even without a fresh crawl, provided the page was indexed clearly enough to be understood. Google AI Overviews tends to draw from pages that already rank well in classic search, so a strong FAQ section sitting on a page with weak traditional SEO may still struggle to surface there.
The anatomy of an FAQ answer AI engines will actually lift
Four habits separate FAQ answers that get lifted into AI responses from FAQ answers that get skipped. None of them require new tools. All of them require rewriting the way most FAQ sections get written.
Write the question the way people talk to AI, not the way they used to type into Google
Google-era FAQ questions were often written to match search queries: short fragments, stripped of natural phrasing, built to catch a keyword. ChatGPT and Perplexity users type full questions, the way they would ask a person. “How much does AEO cost” targets a keyword. “How much does AEO cost for a small business” is closer to what a buyer actually types into ChatGPT.
Pull your questions from real buyer language: customer emails, sales call transcripts, support tickets, and the actual prompts people type into AI tools while researching your category. If you are unsure where to start, read your own support inbox for a week. The wording there is closer to how a buyer will phrase a question to an AI model than anything a keyword tool will produce.
Answer the question in the first sentence, then stop
An AI-citable answer states the fact in the first sentence. Everything after that sentence is optional elaboration, not the answer itself. Here is the weak version of an answer to “does FAQ schema still matter for SEO”: “FAQ schema is a complex topic that touches on structured data, search visibility, and the changing nature of how Google displays results, so the honest answer depends on several factors.” Here is the version an AI model can lift whole: “Yes. FAQ schema does not produce a Google rich result for most sites anymore, but it still helps AI engines cite your content directly.”
The second version can be lifted into a ChatGPT response as written. The first version forces the model to extract the actual point buried inside a sentence built to sound thorough rather than to answer anything.
Keep each answer self-contained
Every FAQ answer should make sense on its own, with no reference to a previous question, a figure above it, or a pronoun with an unclear antecedent. AI models frequently extract a single question and answer pair without the surrounding context of your page. If your answer to question four says “as mentioned above,” the model has nothing above to point to once it lifts that answer out of your page.
Write each answer as if it will be the only sentence anyone ever reads from your entire site. Spell out the subject by name instead of writing “it helps with this.” Write “FAQPage schema helps AI models extract your answer directly” instead.
Two to four sentences, not two paragraphs
The best-performing FAQ answers run two to four sentences. Long enough to answer the question completely. Short enough that an AI model can lift the entire thing without needing to summarize or trim it. If your answer needs six sentences to say what a competitor says in three, you have added information that belongs in the article body, not the FAQ.
Save nuance, caveats, and edge cases for the main article content, then let the FAQ answer stand as the clean, quotable version of the argument you make in full elsewhere on the page.
Matching FAQ answers to your brand voice without losing citability
A direct, two-to-four-sentence answer does not have to read like a form letter. Brand voice and citability are not in conflict, as long as the direct answer still comes first. State the plain fact in sentence one, then let personality show in sentence two or three if your brand relies on personality to build trust.
Consider two versions of an answer to “do you offer a free trial.” A stripped, generic version: “Yes, a free trial is available.” A version with the direct answer plus voice: “Yes. You get fourteen days to test the full platform before any card is charged, and our team walks you through setup on day one.” Both open with a clear yes. The second gives an AI model the same clean fact to lift, with the detail a reader would actually want attached.
The rule to hold onto: personality belongs after the answer, never instead of it. If your brand voice guide pushes toward long, conversational openings before the point lands, the FAQ section is the one place on your site where that instinct should lose.
A worked example: rewriting a real FAQ answer for AI citation
Here is a before and after for a question we see on nearly every AEO client’s site: “how long does it take to see AEO results.”
Before: “AEO results vary significantly based on a wide range of factors including your industry, current site structure, competitive landscape, and the resources you are able to dedicate to the effort, so it is difficult to give a precise timeline without a full assessment.”
After: “Foundational AEO changes show measurable movement in four to eight weeks. Consistent AI citations for competitive queries typically take two to six months, depending on your starting entity authority and content depth.”
The rewrite keeps the honest caveat, timelines depend on starting position, but states a real range first instead of hiding behind “it depends.” An AI model can lift the second version whole. It has nowhere useful to grab in the first.
Here is a second pair, this time for a yes-or-no question: “can I switch AEO providers if I already have an in-house SEO team.”
Before: “This is a common question we receive, and the answer really depends on your specific setup, team structure, and how your existing SEO processes are configured, so we typically recommend a discovery call to walk through the details.”
After: “Yes. AEO work runs alongside an in-house SEO team rather than replacing it, since the two disciplines rely on different signals and measurement. Most clients keep their SEO team focused on rankings and content while AEO work covers schema, entity building, and AI citation tracking.”
How to write FAQ content for AI citation, step by step
Follow these six steps in order. The full sequence takes an afternoon for a single page and becomes routine after the first few.
- Mine the actual questions your buyers ask. Pull from support tickets, sales call transcripts, live chat logs, and the “People also ask” box for your target keyword. Cross-reference against the questions you would type into ChatGPT if you were the buyer, not the questions a keyword tool suggests.
- Draft the direct answer before you draft anything else. Write the one or two sentence answer first, before any supporting explanation. If you cannot state the answer in one sentence, you do not understand the topic well enough to write the FAQ item yet.
- Cut every answer down to two to four sentences. Remove hedges, throat-clearing, and anything that repeats a point already made in the question. Read the answer alone, outside the page, and confirm it still makes complete sense.
- Match your FAQPage schema exactly to the visible page content. Copy the visible question and answer text directly into the schema. Do not paraphrase, shorten, or add anything to the schema that is not also visible on the page.
- Confirm AI crawlers can see the section without executing JavaScript. View the page source, not the rendered DOM, and check that the questions and answers appear in the raw HTML. Server render the section, or build it as static HTML, if it only appears after a script runs.
- Validate the schema, then test it against a real AI query. Run the page through Google’s Rich Results Test to confirm the schema is syntactically valid. Then ask ChatGPT and Perplexity the exact question from your FAQ item and check whether your answer, or a close version of it, appears in the response.
How to pressure-test an FAQ answer before you publish it
Three quick tests catch most FAQ answers that look finished on the page but will not get cited.
The extraction test. Copy only the answer text, paste it into a blank document with no surrounding context, and read it as a stranger would. If a sentence references something outside that blank document, the answer is not self-contained yet.
The first-sentence test. Read only the first sentence of the answer. If that sentence alone would satisfy someone who asked the question out loud, the answer passes. If the first sentence is a lead-in that requires a second sentence to actually answer anything, rewrite it.
The live-query test. Ask ChatGPT or Perplexity the exact question, worded the way your FAQ item is worded. Compare the response to your published answer. A close match confirms the format is working. No match at all after several attempts usually points to a crawler access issue or a schema mismatch, rather than a writing problem.
Where to put your FAQ section
Two placement models work: a dedicated FAQ page and an FAQ section embedded at the bottom of a relevant article or service page. For AI citation, the embedded section usually outperforms the standalone page, because it sits inside content that already establishes context and supporting detail the model can draw on if it wants more than the short answer.
A dedicated FAQ or help center page still has a place for high-traffic support questions unrelated to a specific piece of content, such as billing or account questions. For anything tied to a topic you want AI models to cite you on, embed the FAQ section directly inside the article covering that topic. Our AEO content checklist treats an embedded FAQ section as one of the required elements on every published article, not an optional add-on.
Service and product pages benefit from a shorter, more targeted FAQ section than long-form articles, typically the five or six questions a buyer asks right before deciding, such as pricing, timeline, and how the service differs from adjacent options. Save the broader, more exploratory questions for the article that covers the topic in full, and link between the two so a reader or an AI model following the trail lands on progressively more detailed answers.
How many questions belong in one FAQ section
Five to eight questions per page is the range that performs best across the sites we have audited. Fewer than five and you have not covered enough ground to be useful. More than eight and you are usually padding the section with variations of the same question, which dilutes rather than strengthens your citation odds.
Prioritize breadth over repetition. Eight distinct questions covering different angles of a topic beat eight versions of the same question reworded. If two FAQ items could share one answer, combine them into one.
Should you keep implementing FAQPage schema if Google will not show it?
Yes, for three reasons that have nothing to do with the old rich result. First, AI citation impact does not depend on Google’s rich result eligibility at all. ChatGPT and Perplexity read your schema and your page content directly, independent of what Google chooses to render in its own search results. Second, the small set of sites still eligible for the rich result, mostly government and health domains, proves the format was never fully retired, only restricted. Third, removing schema you already built has a real cost with no offsetting benefit. There is no scenario where dropping valid FAQPage schema helps your AI visibility.
The only mistake worth avoiding is treating FAQPage schema as the whole job. Schema without a well-written, self-contained visible answer accomplishes very little. The schema is the wrapper. The two-to-four-sentence answer inside it is the actual content doing the work.
Some teams go further and strip FAQPage schema entirely after noticing the rich result vanished, treating the markup as dead weight. That decision throws away a working AI citation signal to reclaim a few kilobytes of code. If your development team suggests removing FAQ schema during a cleanup pass, this is the one section worth pushing back on.
FAQ rich results versus FAQ content for AI citation
The mechanics changed enough that a side-by-side comparison is worth spelling out.
| Dimension | FAQ rich results (Google, pre-2023) | FAQ content for AI citation (now) |
|---|---|---|
| Primary reader | A person scanning search results | An AI model generating an answer |
| Where it appears | Expandable accordion in the SERP | Lifted or summarized inside a chat response |
| What triggers it | Valid FAQPage schema, on nearly any site | Clear question phrasing, a self-contained answer, and matching schema |
| What matters most | Schema validity alone | Schema validity plus the quality of the visible answer |
| Current status | Restricted to a small set of government and health sites since 2023 | Active and growing across ChatGPT, Perplexity, Google AI Overviews, and Copilot |
Keep FAQ content current
AI models deprioritize outdated information, and FAQ answers age quietly. A pricing answer written two years ago, a step-by-step answer describing a tool that changed its interface, or a figure that has since moved can all quietly turn a once-citable answer into one an AI model will skip in favor of a fresher source.
Review your highest-traffic FAQ sections every quarter. Update numbers, confirm tool references still match the current product, and reread each answer as if you were a stranger encountering it for the first time. If a competitor’s more recent answer reads as more current than yours, an AI model will notice the same gap you would.
Set a recurring reminder rather than relying on memory. A quarterly calendar entry tied to your highest-traffic pages costs a few minutes and catches the pricing figure, the tool name, or the detail that quietly went stale since the last review.
Common mistakes that keep FAQ content from getting cited
Six mistakes account for most of the FAQ sections we see that never get cited despite looking complete on the page.
- Writing questions for keywords instead of people. A fragment built to match a search term rarely matches how someone phrases a question to an AI model.
- Padding answers with hedges and throat-clearing. Qualifiers like “it depends” and “in many cases” push the actual answer further from the first sentence, where AI models look first.
- Letting schema drift from the visible text. Any mismatch between what the page shows and what the schema states undermines both.
- Hiding the FAQ behind a script that never renders in raw HTML. A crawler that cannot see the content cannot cite it, regardless of how the answer reads to a human visitor.
- Answering a different question than the one asked. If the question is “how much does X cost” and the answer opens with a paragraph about what X includes, the model has to work to find the number.
- Recycling the same five generic questions on every page. Generic questions like “what is AEO” belong on one definitive page. Repeating them everywhere dilutes the specific, page-level questions that would actually earn a citation for that page’s topic.
Most of these mistakes come from writing FAQ content the way FAQ content used to be written, optimized for a rich result that Google will not show most sites anymore. Fixing them is mechanical work once you know what to look for, which is exactly why FAQ sections remain one of the fastest wins available in AEO.
FAQ content is one lever, not the whole program
FAQ sections are one of the fastest levers available in AEO, but they do not replace the rest of the work. A page with a flawless FAQ section and no entity authority, no technical foundation, and no answer-first structure in the main body will still struggle to get cited consistently. Our full playbook on getting your brand cited by ChatGPT covers the levers that work alongside FAQ content, including source authority and entity clarity.
The pattern holds across every audit we run. Pages with a strong FAQ section and weak entity signals see occasional citations. Pages with both a strong FAQ section and strong entity signals see citations that hold steady month over month. FAQ writing is the fast half of the equation. Entity building is the slow half that makes the fast half durable.
Treat FAQ writing as the fastest fix available today, not the only fix your content needs.
Getting an outside read on your FAQ content
If you want an outside read on whether your existing FAQ sections are actually structured for AI citation, that is exactly the audit we run inside our AI Visibility and AEO service. We check schema validity, crawler access, answer structure, and run live citation tests across ChatGPT, Perplexity, and Google AI Overviews, then hand you a prioritized list of which FAQ sections to fix first.
