All of the Lab Journal
SEO Writing

How to Write FAQs That Land in AI Answers

Baptiste Lacroix
Founder of MentionLab
BlueWritten with Blue
July 15, 2026

Write FAQs by pulling real questions from your support inbox, sales calls, and search data, then answering each one in 40 to 90 words that stand on their own. The self-contained part is what decides whether an AI answer can lift your response or has to skip it. A question-and-answer block is the closest thing on your website to the shape a model actually retrieves.

FAQs have spent the last decade being written as a support deflection tool, something you publish to cut down on repetitive tickets. That's still a fine side effect. But the search results and AI answers around "how to write faqs" now reward something else: a block of real questions and real answers that a model can lift wholesale and attribute to your site.

What Makes an FAQ Answer Get Picked Up by an AI Answer?

A question-and-answer pair that stands on its own is already shaped like what an answer engine retrieves: an explicit question, a short and complete response, no dependency on the paragraph above it. That structural match is why FAQ content punches above its weight in AI answers, even when nobody writes it with that in mind.

The stakes keep rising. Pages with an AI-generated summary convert to a click only 8% of the time, versus 15% without one, roughly half as often (Pew Research Center, July 2025). Even a page cited as a source inside an AI summary earns a click only 1% of the time (Pew Research Center, July 2025). Before writing an FAQ, it helps to understand what zero-click search means for your traffic and how pages get cited in AI Overviews.

Where Do You Get the Questions Worth Answering?

The best FAQ questions come from actual demand: your support inbox, sales call notes, and your site's internal search logs, not a brainstorm around a conference table. Google's own People Also Ask box is a fourth source worth checking, since it shows the exact phrasing real searchers use for your topic.

The trap almost every FAQ falls into is answering questions that are easy to answer, not the ones people actually ask. Teams happily write three paragraphs on a policy nobody asks about because it's a clean topic to cover, while the question customers raise every week goes unwritten because it's messier to answer. An FAQ built from convenient questions never gets picked up by search or by AI answers, for the same reason: nobody types that question into either one. Once the real questions are sourced, optimizing content for answer engines covers how the rest of the page should support the FAQ block.

How Do You Write an Answer a Model Can Actually Lift?

Answer the question in the first sentence, keep the full answer between 40 and 90 words, and write it so it still makes sense if lifted out of the page and dropped into a chat window. That self-containment is the one requirement most FAQ advice skips.

Six things to check on every answer:

  1. Put the direct answer in the first sentence, not the third.
  2. Open yes or no questions with an explicit yes or no.
  3. Cut phrases that depend on context, like "as mentioned above."
  4. Name the entity instead of a pronoun: "MentionLab publishes..." not "It publishes..."
  5. Answer fully instead of pointing to another page for the real answer.
  6. Write the question the way a reader actually types it.

The gap between an extractable and a non-extractable answer is usually small on the page and large in practice. Not extractable: "It depends on which plan you picked and what got set up during onboarding." Extractable: "The Basic plan includes 10 hours of support a month; the Pro plan raises that to 25 hours." The first version needs missing context to mean anything; the second stands on its own.

Putting the answer in the first sentence is the same BLUF principle applied at sentence level, paired with turning sentences into facts machines can reuse once you name entities instead of leaning on pronouns. It's close to what we check when structuring the Q&A blocks inside MentionLab's own blog articles before they ship.

This article, Blue could have written it for you: content optimized for Google + AI, without you writing a single word.

Try mentionLAB

Does FAQ Schema Still Do Anything After Google Retired FAQ Rich Results?

Yes, but not for the reason the last few years trained you to expect. Google retired FAQ rich results from Search starting May 7, 2026, following an August 2023 restriction limiting the rich result to well-known government and health sites (Google Search Central, developers.google.com). Rich result reporting and testing tools followed in June 2026, and Search Console API support for the rich result drops in August 2026 (Search Engine Land, May 2026). Google retired FAQ rich results in May 2026, so FAQPage markup on a typical business site now produces nothing visible in Google Search.

That doesn't make the markup pointless. FAQPage remains a valid, non-deprecated schema.org type, still defined as a WebPage presenting frequently asked questions and used across an estimated 1 million to 10 million domains (schema.org). Google itself says unused structured data causes no problems and has no visible effect (Google Search Central, developers.google.com), so leaving the markup in place costs nothing. It now mirrors your visible text for machines that still parse it: a poorly written FAQ, tagged perfectly, is still a poorly written FAQ. For how to implement FAQPage markup, see the full setup.

When Should You Not Write an FAQ?

Don't write an FAQ when the answer already belongs in your main content. A catch-all FAQ page that repeats what your site says elsewhere creates duplicate content and puts two of your own pages in competition for the same question.

This objection isn't new. The UK's Government Digital Service argued in 2013 that FAQs are convenient for the writer and expensive for the reader, since they let teams skip frontloading the answer where it belongs and dump it into a separate list instead (Government Digital Service, 2013). That argument holds only against one shape of FAQ: the sitewide catch-all page duplicating content that lives somewhere else. It doesn't hold against a short Q&A block placed on the page that already covers the topic, answering the exact follow-up question a reader, or a model, would have next.

The actionable version: keep FAQs attached to the page about the subject, not filed in a central dump. If an answer deserves its own H2, it's not a FAQ entry, it's a section, matching your format to what the SERP actually rewards rather than defaulting to a Q&A list out of habit.

Frequently Asked Questions

What is the format for FAQ?

The format is a question written exactly as a reader would ask it, followed by a short, complete answer. Put the question in a heading, never disguised as a marketing statement with a question mark tacked on. Keep one question per answer instead of stacking two or three under a single heading, since that breaks a model's ability to extract a clean pair.

What should be included in FAQ?

Include the questions your team actually receives, one clear answer per question, and a way to reach a human when the answer isn't enough. Leave out questions that are convenient to answer but nobody asks, since padding an FAQ with easy filler dilutes the questions real visitors and real search queries are actually asking.

What does a good FAQ look like?

A good FAQ is short, made of real questions grouped by topic, with every answer understandable on its own. The simplest test: pull one answer out of the page and read it in isolation. If it still makes sense without the surrounding questions, it's doing its job; if it depends on another answer's context, it needs rewriting.

What are FAQs examples?

Common FAQ categories include shipping and delivery timelines, returns and refunds, pricing and billing, eligibility requirements, and account setup. The selection criterion matters more than the category list: include a question only if it comes from a real support ticket, sales call, or search query, not because a competitor happens to cover it.

Is it FAQ or FAQs?

Both are correct, and they mean different things. "FAQ" refers to the page or section itself, frequently asked questions, while "FAQs" refers to the individual questions it contains. Common US usage treats "FAQ" as the label for the whole page, as in "check the FAQ," while "FAQs" lists the questions inside it.

Blue handles your SEO and your GEO. On autopilot.

You approve, she produces content optimized for Google + AI.

Join the Lab · 5-day trial

Read next

SEO Writing

What Is AI Slop and How to Keep It Off Your Blog

AI slop is the term for low-quality content produced in volume with generative AI, built with little regard for accuracy, relevance, or whether anyone asked for it. Merriam-Webster named it 2025's word of the year, but the label now gets thrown at almost anything AI touches, which makes it a judgment call more than a technical category. This article defines the term, shows what it actually looks like in text (not just glitchy AI images), and lays out a four-part test, drawn from criteria that Google, Wikipedia, and independent researchers already use, to tell slop apart from useful AI-assisted writing.

SEO Writing

How to Write a Listicle That Search and AI Both Pick Up

Most advice on writing a listicle skips the one number that actually matters: how many items your target search result page rewards. That number isn't a rule of thumb, it's measurable, and on the keyword "how to write a listicle" it measures out to seven. Google's own featured snippet proves it: the ranked page it pulls from publishes ten items, but the snippet displays only seven, and the cut lands exactly on the median across all six ranked how-to pages. This article covers the method: count the items on your top ranked pages, take the median, check where the current snippet cuts, then front-load your strongest items inside that visible ceiling before you publish a single word.

SEO Writing

Article Schema for Blog Posts That Machines Can Parse

Article schema is a JSON-LD block that tells Google and AI systems the headline, author, and publish date of a blog post instead of leaving that information to be inferred from the page layout. This article resolves the Article vs. BlogPosting vs. NewsArticle decision with a simple table, hands over a complete copy-paste JSON-LD template with every recommended property filled in, and explains a gap most explainers skip entirely: how AI crawlers like GPTBot and PerplexityBot lean on the same fields differently than Googlebot does when deciding whether a page is citable. A validated properties table, a common-mistakes list drawn from Google's own documentation, and a three-step validation walkthrough round out the piece.