An FAQ page for AI search in 2026 is a set of real customer questions, each answered in a self contained 40 to 100 word block that resolves the question in its first sentence, names specific entities and numbers, and sits on the service page it relates to, wrapped in FAQPage schema. Google removed FAQ rich results from its search results on May 7, 2026, but the schema still helps ChatGPT, Perplexity, Claude, and Google AI Mode parse question and answer pairs, and Kevin Indig’s Growth Memo analysis of 18,012 ChatGPT citations found that 44.2% of quoted passages come from the first 30% of a page. Build the FAQ that way and you give every answer engine a clean chunk to quote.
Which questions belong on an FAQ page built for AI search?
The questions your buyers type into ChatGPT and Google, in their words, not yours. Pull them from four sources, in this order of value: Google Search Console, sales call transcripts, People Also Ask (plus AlsoAsked and AnswerThePublic), then Reddit. The overlap between them is your shortlist.
Two numbers frame why selection matters more than volume. Semrush’s 2026 AI Visibility Index, built on 126 million US prompts from January through April 2026, found ChatGPT cites an average of 15 sources per response while Gemini cites 3. Previsible’s July 2026 State of AI Discovery report found ChatGPT carries 92.4% of trackable standalone LLM referral traffic, with Claude growing 64x over the tracked period. The engines have room for you, but only on questions where your page is the best passage.
1. Google Search Console queries
Open Performance, filter to the service page, and sort by impressions. Any query phrased as a question (“how much does,” “what happens if,” “can I”) with impressions but no clicks is a question you are being considered for and losing. Those go to the top of the list. Export 16 months so seasonal questions show up.
2. Sales call and intake transcripts
Search your last 20 discovery call transcripts for question marks. The questions people ask a human right before buying are the questions they ask ChatGPT right before calling, and the exact wording from a call (“do I have to be in the office to sign”) usually beats the polished version.
3. People Also Ask, AlsoAsked, and AnswerThePublic
Expand Google’s People Also Ask box four levels deep for your head term. AlsoAsked maps the same tree in one click; AnswerThePublic groups the prepositional variants (versus, near, without, for). Keep the questions where you have a specific answer. Drop the “it depends” ones unless you can say what it depends on.
4. Reddit and community threads
Profound’s citation analysis puts Reddit at 1.8% of ChatGPT citations and 6.6% of Perplexity citations, so the engines already read those threads. Search your niche subreddit for questions with 20 or more comments and no clear winner. Answer those and you are competing with a thread, not a marketing team.
Not sure whether ChatGPT and Google AI Mode are quoting your FAQ answers or a competitor’s? Get your free AI visibility audit and see which of your questions are earning citations right now.
How long should each FAQ answer be, and what goes in the first sentence?
40 to 100 words, and the first sentence must answer the question on its own. Contently’s March 2026 review of FAQs that LLMs cite puts the sweet spot at 40 to 80 words in two to four sentences, and our own client audits land in the same band. Under 40 words reads as thin and rarely gets quoted. Over 100 words, the engine chunks your answer mid thought and a shorter competitor takes the citation.
Structure every answer the same way:
- Direct answer sentence. Restate the subject and resolve it. “A first DUI in South Carolina carries a 48 hour to 30 day jail sentence and a $400 fine” beats “Several factors determine sentencing.”
- One sentence of specifics. A number, a named entity, a date, a jurisdiction, a product name.
- One sentence of context or exception. The edge case, the “unless,” the next step.
- No links inside the answer. Links inside the answer give the engine a reason to cite the link target instead of you. Put them after the block.
Here is a worked example for a residential HVAC company in Phoenix:
How much does it cost to replace a central AC unit in Phoenix?
Replacing a central air conditioner in Phoenix costs $6,500 to $12,800 installed for a 3 to 5 ton, 15.2 SEER2 system in 2026, according to our last 140 installs. Carrier and Trane units sit at the top of that range; Goodman and Rheem at the bottom. Permits from the City of Phoenix add $150 to $300, and APS rebates return up to $1,125 on qualifying high efficiency models.
That is 62 words. It answers in the first sentence, names five entities (Carrier, Trane, Goodman, Rheem, APS), gives four numbers, states the year, and stands alone with zero context from the rest of the page. Compare it to what most sites publish (“costs vary depending on the size of your home and other factors, contact us for a quote”): zero entities, zero numbers, nothing to quote.
Does FAQPage schema still matter after Google removed FAQ rich results?
Yes, for extraction rather than for a SERP feature. Google added a deprecation notice to its FAQ structured data documentation on May 7, 2026, the dropdowns disappeared the same day, the FAQ search appearance filter and Rich Results Test support were removed in June, and Search Console API access for FAQ data ended in August. Google’s note said it would continue using FAQ structured data to understand pages, and FAQPage remains a valid schema.org type.
What changed is the payoff. The rich result was a click through rate feature with a fixed ceiling. FAQPage schema now labels each question and answer pair so a crawler does not have to guess where one answer ends and the next begins, which helps any system that chunks pages into passages before retrieval: ChatGPT search, Perplexity, Bing Copilot, and Google AI Mode. Google Search Central’s “AI features and your website” documentation says no special structured data is required for AI Overviews or AI Mode, and that structured data helps Google understand content. Treat the schema as a parsing aid, not a ranking lever.
Implementation rules that still apply:
- Use JSON-LD, one FAQPage object per page, each Question with exactly one acceptedAnswer.
- Schema answer text must match the visible text word for word. Marking up hidden or different content violates Google’s structured data guidelines.
- Validate at validator.schema.org, since the Rich Results Test no longer reports FAQ.
If you want the legal specific version of this argument, why every law firm needs an FAQ page covers the case for practice area pages in detail.
Should the FAQ live on one hub page or on each service page?
On each service page, with a hub only as a secondary index. Engines retrieve passages by topical match, so a “how much does Invisalign cost in Austin” answer on your /invisalign/ page beats the same answer buried on a 60 question /faq/ page. Kevin Indig’s finding that 44.2% of ChatGPT citations come from the first 30% of a page argues for the same placement: question 50 on a hub page sits in the part of the document engines rarely quote.
The working pattern for most service businesses:
- Service page: 6 to 10 questions specific to that service and location, after the main body copy, under an H2 like “Common questions about [service] in [city].”
- Pricing or fees page: the cross cutting money questions (payment plans, insurance, deposits, refunds).
- Hub /faq/ page (optional): a short index linking to each service page’s FAQ block. Duplicated answers split the citation between two URLs and neither wins.
One more rule from Google’s guidelines: FAQPage is for questions the site itself answers. User submitted Q&A belongs under QAPage.
How do you keep an FAQ page fresh enough for ChatGPT to keep citing it?
Update it on a schedule and prove the update in the markup. Ahrefs’ analysis of 17 million ChatGPT citations found 76.4% came from pages published or updated within the previous 30 days, the strongest recency preference of any engine it measured. An FAQ block with 2024 prices is losing to a thinner competitor page dated last month.
Freshness is a workflow:
- Quarterly review. Reopen the Search Console export, add the new questions, retire the ones with no impressions, and update any number that moved (prices, timelines, statutes, product versions).
- dateModified in the schema. Add it to the page’s Article or WebPage schema and update it only when content changes. Bumping the date without changing text is a pattern Google’s spam policies call out.
- Visible “Last updated” line directly above the FAQ block.
- Internal links into the FAQ block. Link from blog posts and related service pages to the specific question with an anchor (/invisalign/#cost). Internal links tell the crawler which passages matter.
The retrieval step where chunking happens is covered in how ChatGPT search actually works.
How do you measure whether AI engines are quoting your FAQ answers?
Combine one free first party report, one paid prompt tracker, and a manual spot check. No single source shows every engine.
Bing Webmaster Tools AI Performance report. Microsoft launched it on February 11, 2026. It shows AI impressions, AI clicks, and AI click through rate for Copilot and Bing’s AI summaries, plus the grounding queries Copilot generated internally to retrieve your page. Since ChatGPT search draws on Bing’s index, those grounding queries are the closest free view of which FAQ questions match real retrieval.
Google Search Console. There is still no AI Overview or AI Mode filter; AI features traffic is folded into the Web search type. Watch impressions for question phrased queries on the FAQ page. Impressions rising while clicks stay flat is the footprint of being quoted in an AI answer.
Prompt trackers: Otterly.ai, Peec AI, and Profound. These run a fixed prompt set through ChatGPT, Perplexity, Google AI Overviews, and AI Mode on a schedule and record which URLs get cited. Otterly.ai publishes self serve tiers at $29, $189, and $489 per month. Peec AI runs €89 to €199 per month. Profound is enterprise priced through a demo. Load your FAQ questions as the prompt set, verbatim, and track citation share per question.
Manual spot check. Monthly, ask ChatGPT, Perplexity, Claude, and Google AI Mode your 10 highest value questions and record whether your domain appears in the sources.
If a competitor is being quoted on a question you answer better, the cause is almost always one of three: your answer runs past 100 words, your first sentence does not resolve the question, or the page is not in the engine’s index. How to get cited by ChatGPT covers the index and crawler side of that diagnosis.
Frequently asked questions
How many questions should an FAQ page have for AI search?
Six to ten per service page. Fewer than six leaves obvious buyer questions unanswered, and more than ten pushes later questions past the first 30% of the page where Kevin Indig’s Growth Memo data shows most ChatGPT citations originate. If you have 25 legitimate questions for one service, split them across the service page, a pricing page, and a process page rather than stacking them on one URL.
Should I remove FAQPage schema now that Google dropped FAQ rich results?
No. Google removed the visual rich result on May 7, 2026, but its deprecation notice states it will keep using FAQ structured data to understand pages, and FAQPage remains a valid schema.org type. The markup labels question and answer pairs for any crawler that chunks pages into passages, including ChatGPT search, Perplexity, and Bing Copilot. Keep it, validate it at validator.schema.org, and make sure the schema text matches the visible text.
What is the ideal FAQ answer length for ChatGPT and Perplexity?
40 to 100 words, with the question resolved in the first sentence. Contently’s 2026 analysis of FAQs that LLMs cite places the peak at 40 to 80 words in two to four sentences. Answers under 40 words look thin and rarely get quoted; answers over 100 words get split mid passage. Include at least one number and one named entity so the passage reads as evidence rather than opinion.
Can I use ChatGPT to write the FAQ answers?
Yes for the draft, no for the final. AI drafted answers tend to produce the “it depends on several factors” pattern that engines skip. Have the practitioner add the specifics: your actual price range, your city’s permit fee, the statute number, the product version. Specificity is what earns the citation. Then run the answer through the 40 to 100 word check and confirm the first sentence answers the question on its own.
How long before an FAQ page starts getting cited?
For an established domain with healthy crawl activity, 30 to 90 days after publishing, and faster on Bing Copilot and Perplexity than on ChatGPT. Ahrefs’ 17 million citation study found 76.4% of ChatGPT citations point to pages updated in the last 30 days, so a refresh cadence matters as much as the launch. Newer domains take longer because retrieval systems weight domain trust and third party mentions heavily.
Does an FAQ page help with Google AI Overviews as well as ChatGPT?
Yes. Google Search Central’s “AI features and your website” documentation says pages need only be indexed and eligible for a snippet to appear in AI Overviews and AI Mode, and that structured data helps Google understand content. Previsible’s 2026 report found AI discovery inside Google (AI Overviews plus AI Mode) drives more AI influenced traffic than every standalone LLM combined, so a well built FAQ page earns on both surfaces.
The bottom line
The FAQ block is the smallest unit of content an AI engine can quote, and it is the one most sites still get wrong. Google removing the FAQ dropdown took away the vanity reward and left the real one: 40 to 100 word answers that resolve a real buyer question in the first sentence with named entities and numbers, placed on the service page they belong to, marked up in FAQPage schema, and refreshed inside the 30 day window ChatGPT favors. Ship that on your top three service pages this month, load the same questions into a prompt tracker, and within 90 days you will know which answers the engines quote and which a competitor still owns.
Want to know which of your FAQ answers ChatGPT, Perplexity, and Google AI Mode are quoting today, and which questions send buyers to someone else? Run your free AI visibility audit and get the query by query breakdown.
Tagged