How to write an FAQ in 2026: what 1,733 AI answers and 1,641 Google questions show

Last updated November 12, 2026 · ~8 min read

The short answer: Collect the questions buyers really ask, write each one in their words, and answer it in the first sentence with yes or no, a number or a name; then add a fact that can be checked. Put each answer on the page it is about, include the doubts buyers raise, and keep it current. In our data, only 2.8% of 1,733 AI answers to buying questions cited an FAQ page, and those that did were mostly about one question or one topic. Google now writes 96% of the answers in its People also ask box itself, and it stopped showing FAQ dropdowns on May 7, 2026.

How to write an FAQ in 7 steps

1. Collect the questions people really ask. Read your support inbox, chat logs, sales call notes and reviews. Then look at the “People also ask” box in Google and at what AI assistants say when someone asks about your kind of product. Keep the questions that come up more than once.
2. Write each question in the buyer’s words. “Can I take these on a plane?”, not “Travel compatibility”. One question per heading.
3. Answer in the first sentence. Start with yes or no, a number, a name or a date. A third of the questions Google shows buyers can be answered yes or no.
4. Back the answer with a fact someone can check. A size, a price, a policy, a standard, a date. Link to the page where it is written.
5. Include the uncomfortable questions. Downsides, “is it worth it”, why people switch away. Buyers ask them, and someone else will answer them if you don’t.
6. Put each answer where it belongs. Product questions on the product page, price questions on the pricing page, delivery and returns on the policy pages. If you keep a separate FAQ page, group it by topic.
7. Keep it current. Change the answer when the price, policy or product changes, and show the date it was last checked.

Most FAQ guides stop at steps like these. We wanted to know how they hold up now that AI assistants answer a lot of buyers’ questions, so we went back to the answers we had already collected.

What our data shows: AI answers rarely cite FAQ pages

We looked at 1,733 answers from ChatGPT, Perplexity and Google’s AI Overviews to 120 US buying questions: 60 shopping questions and 60 B2B ones, asked several times between October 22 and November 9, 2026. Together they cited 13,494 pages. We counted a page as an FAQ page when its address or title said FAQ or “frequently asked questions”.

Only 49 of the 1,733 answers (2.8%) cited an FAQ page, and FAQ pages made up 0.4% of all the pages cited. Most of what got cited were articles, product pages, comparison pages and reviews.

That does not make FAQs useless. It means a page called “FAQ” is rarely the page an AI answer picks. The ones that were picked had something in common.

The FAQ pages that did get cited were about one thing

The 49 answers cited 19 different FAQ pages. 16 of them covered one question or one topic; only 3 were a company’s general FAQ.

• One question per page (5 pages). Mayo Clinic’s “Dental floss vs. water flosser: Which is better?”, cited by all three engines; TP-Link’s “How many Deco units do I need?”; Google’s “What is a Chromebook good for?”.
• One topic per page (11 pages). ENERGY STAR’s smart thermostat FAQs, Trane’s “14 Smart Thermostat FAQs, Answered”, ZoomInfo’s pricing FAQs, an accounting firm’s SOC 2 FAQ.
• A general company FAQ (3 pages), such as Ray-Ban’s customer FAQ page.

Five of the 19 came from a government agency or a medical or professional body. Nineteen pages is a small number, so read this as a direction, not a rule: an answer cites the page that answers the question, and a page about one question or one topic is more often that page than a long list.

Google’s AI Overviews answer from forum threads instead

Six in ten AI Overviews cited Reddit, Quora or Stack Exchange; ChatGPT almost never did. Those threads are other buyers asking the same questions your FAQ should answer, in their own words, and other buyers answering them. If your answer is missing, vague or hidden behind “contact us”, the thread’s answer is the one people read. Which sites AI Overviews cite, by question, is in our AI Overview sources study.

Many cited pages were written as the answer to one question. Leaving forums aside, 27% of the pages AI Overviews cited had a question as their title, such as “Is SOC 2 difficult?”, against 18% for Perplexity and 6% for ChatGPT.

“People also ask” is a free list of buyer questions, but Google writes the answers now

The same Google searches showed a People also ask box in 413 of 597 cases: 80% of shopping searches and 53% of B2B ones, 1,641 questions in all. That is the best free list of follow-up questions you will find. But look at who answers them:

• 1,577 of the 1,641 answers (96%) were written by Google’s AI, as a small AI Overview under the question. Only 4 quoted a single page; the other 60 had no answer in our data.
• So the old trick of writing an FAQ to “win” a People also ask box is mostly gone. Use the box to find questions, then answer them well on your own pages.

Here is how those 1,641 questions start, which tells you how your answers should start:

Two more things stood out:

• About 1 in 13 questions was a doubt: a downside, a risk, “is it worth it” or why people leave. “What is the downside of a Chromebook?”, “Is there a downside to using an air purifier?”, “What are the pros and cons of using NetSuite ERP?”. An FAQ that only answers the easy questions skips the ones buyers ask right before they decide.
• B2B buyers ask about price more: 8% of B2B questions were about price or cost, against 3% for shopping. If you sell to businesses and your price is “contact sales”, at least say what the price depends on and where it starts.

The box does not show up for every search, and neither does the AI Overview above it; how often each appears, and why, is in Google AI Overview not showing up.

FAQ schema in 2026: no more dropdowns in Google

For years, FAQ pages came with FAQPage markup so Google would show the questions as dropdowns under the result. That has ended. Google’s Search Central changelog says the FAQ rich result stopped appearing in Google Search on May 7, 2026, and Google removed its documentation in June. In our saved searches from October and November, none of the 4,804 regular results carried one.

Google’s page on AI features says there is nothing extra to do to appear in AI Overviews or AI Mode: no special markup, no AI text files, no special schema. It does say that any structured data you keep should match the text people can see on the page.

So: don’t add FAQ markup expecting it to change how you show in Google. If you already have it, it does no harm as long as it matches the visible questions and answers word for word. We did not test whether ChatGPT or Perplexity use FAQ markup; what we can see is the pages they cite.

How to write each answer

One pattern works for almost every question: the answer, then the proof, then the condition.

1. The answer, in a first sentence that still makes sense if it is the only one someone reads.
2. The proof: the fact behind it, with a number or a source.
3. The condition: when the answer changes, or who it is not for.

Two made-up examples, one for a product and one for a B2B service:

The strong answers give the facts AI answers tend to repeat. In our test of 513 AI answers to shopping questions, 79% said who a product is for, 54% gave a measurable figure and 53% a price; more in how to write product descriptions. Facts on your own page also matter because AI answers get them wrong: Hasmord Research’s AI Fact-Checking Test (November 2026) found that all 11 wrong shopping answers it checked linked the seller’s own site, which had the right value.

Where to put your FAQs

• On the product page: the questions about that product, answered for that product. This is where buyers and search engines arrive with the question. See product page SEO.
• On the pricing page: what is included, what costs extra, what the price depends on, contracts and cancelling.
• On policy pages: delivery, returns, warranty, data and security, each answered on its own page.
• A separate FAQ page, if you have one, grouped by topic with a heading per question and a link to the page that has the full answer. Short is fine; a page of 80 questions is hard for anyone to use.
• A page of its own for a big question, when buyers ask it often and the answer takes more than a paragraph, the way Mayo Clinic has one page for “dental floss vs. water flosser”. For B2B, comparison and “how to choose” questions are the usual candidates; what gets cited for each is in B2B content marketing.

Common FAQ mistakes

• Questions nobody asks (“Why is our product so good?”).
• Answers that are sales copy and never say yes, no or a number.
• “Contact us” as the answer to a price question.
• One answer that covers three questions.
• Old answers: last year’s price, a discontinued model, a policy that changed.
• FAQ markup that doesn’t match what the page shows.

Method and limits

120 buying questions we wrote: 60 shopping questions in 12 product categories and 60 B2B questions in 12 categories, listed in our AI Overview sources study and our B2B study. Each was asked for a US location: the shopping questions three times on October 22 and three times on November 5, 2026; the B2B questions three times on November 5 and once on November 9. ChatGPT answers come from chatgpt.com search, not signed in, and count the sources the answer lists; Perplexity counts the pages its answer cites; Google counts the pages listed in the AI Overview, which appeared in 533 of the 597 searches that loaded (DataForSEO). A page counts once per answer. FAQ page = “faq” or “frequently asked” in the address or title; Google hides part of its citation addresses, so for Google we could only go by the title for those. Question title = a title that ends with a question mark or starts with a question word. People also ask = every question in the box on the same searches, counted once per search. The labels for the 19 FAQ pages were read by hand. One market, English only, and questions we chose, so read the numbers as directions. All our numbers are on AI search statistics.

Where Hasmord fits

Hasmord is the AEO-native sales platform. Buyers ask questions across AI assistants, search, and sales conversations. Hasmord uncovers where your answers about your products and services fall short, turns evidence into better answers, finds potential customers for you, and connects every lead to your sales process.

An FAQ is only as good as the questions in it. Hasmord’s first step, which uncovers where your answers about your products and services fall short, starts from your buyers’ questions: AI answers asks them in ChatGPT, Claude, Gemini and Perplexity and keeps every answer and its sources, so you can see which questions you are missing from and who answers them instead.

The second step turns evidence into better answers: Hasmord compares those answers with your product facts and prepares fixes, product by product, for your team to approve. Find customers then finds potential customers already looking for what you sell, with the reason each one fits; only the people you verify join your CRM, and Hasmord never contacts anyone for you. In the CRM, each lead keeps the question that brought it. See pricing.

Frequently asked questions

How do you write an FAQ?

Collect the questions people really ask, from support emails, sales calls, reviews and Google’s People also ask box. Write each question in the buyer’s words, answer it in the first sentence with yes or no, a number or a name, then add a fact that can be checked. Put each answer on the page it is about and keep it current.

What should be in an FAQ?

The questions buyers ask before they decide: what it is, whether it fits their case, what it costs, how long it lasts, and the doubts. In 1,641 People also ask questions on buying searches, 38% started with “what”, 33% could be answered yes or no, and about 1 in 13 was about a downside or whether something is worth it.

What does a good FAQ look like?

One question per heading, in the words buyers use, with the answer in the first sentence, the fact behind it in the second, and when the answer changes in the third. It is grouped by topic, short enough to scan, and placed on the page the question is about.

Do AI answers cite FAQ pages?

Rarely. Hasmord Research’s FAQ Citation Study (November 2026) found that 49 of 1,733 answers from ChatGPT, Perplexity and Google’s AI Overviews to 120 buying questions (2.8%) cited an FAQ page. The FAQ pages that were cited were mostly about one question or one topic, not a company’s general FAQ.

Is FAQ schema still worth it?

Not for Google’s dropdowns: Google stopped showing the FAQ rich result on May 7, 2026, and says no special markup is needed to appear in AI Overviews. Existing FAQ markup does no harm if it matches the visible text, but the questions and answers on the page are what matter.

Should FAQs go on product pages or a separate page?

Both, for different questions. Questions about one product belong on that product’s page, price questions on the pricing page, and delivery or returns on policy pages. A separate FAQ page works as an index grouped by topic, and a question that needs a long answer can have a page of its own.

Related: how to write product descriptions · product page SEO · Google AI Overview not showing up · AI Overview sources study · B2B content marketing · AI search statistics.