All posts

FAQ schema for SaaS: when FAQPage markup helps AI answers and when it is noise

welaunch·November 17, 2026

FAQ schema (FAQPage markup) earns its place when it marks up questions buyers actually type into Google or ask ChatGPT, written in their words, with answers that match the visible page text. It becomes noise when you invent marketing questions nobody searches instead of the ones buyers actually type.

So the work is not "add FAQ schema to every page." The work is picking the right questions first, then marking them up.

1. Confirm what FAQ schema actually is and whether it still does anything

FAQ schema is a block of structured data, usually JSON-LD in the page <head>, that tells search engines "this page contains a list of questions and answers." The type is FAQPage, each item is a Question with an acceptedAnswer. That is all it is: a machine-readable copy of something already on your page.

What changed is what Google does with it. For years an FAQPage block could get you an expanded listing with dropdown questions under your result. Google has pulled back on showing the expanded FAQ listing for most sites, so do not count on one appearing for yours. The markup is not deprecated. It is still a documented type and Google still parses it. It just does not buy most SaaS sites a bigger blue-link result any more. If you want to track whether that changes, Google's Search Central updates log is the one page worth a bookmark.

How to check you got this right: ask yourself what outcome you expect. If the honest answer is "a rich result," stop and reconsider. If it is "a clean, parseable Q&A block that search engines and AI crawlers can lift," carry on.

Common mistake: treating FAQ schema as a ranking lever. It is a formatting signal, not a vote.

2. Pull real buyer questions from Google and AI engines instead of guessing

Open an incognito window. Search the thing your buyer searches. Write down every People Also Ask question, click one to expand, and write down the ones that appear underneath. Do this for your three or four highest-intent terms. Then go to Google Search Console, filter queries by ones containing "how", "what", "can", "vs", "does", and export them.

Then do the part most teams skip: ask the same questions to ChatGPT, Perplexity, Gemini and Google AI Mode, and read what comes back. Note which products get named and which sources the answer leans on.

The reason to run the AI side as well as the Google side is that the answers are less stable than they look. In welaunch's test of live AI answers (27 Sep 2026, 06-aeo-live.md), 434 answers were pulled from ChatGPT, Perplexity, Gemini and Google AI Mode for buyer questions across seven software categories. In the AI-visibility category, Profound was named in 23 of 63 answers and Otterly and Peec in 17 each, according to that same test. One run is an anecdote. Four runs is a pattern.

How to check it worked: you should end up with a short list of questions in the buyer's phrasing, not yours. "Does it integrate with HubSpot" is a buyer question. "Why choose our platform" is not.

Common mistake: writing the FAQ list in a planning doc from memory. Every question you invent is a question nobody asked.

3. Check whether those questions are even the ones AI engines cite answers for

Before you mark anything up, look at what each engine actually cites when it answers your questions. They behave very differently, and that changes how much your own page can matter.

From welaunch's test of 434 live AI answers to buyer questions in seven software categories (27 Sep 2026, US):

EngineAvg sources per answerAnswers with no sourceShare citing reddit.com
ChatGPT5.8 across 110 answers76% (7 of 110)
Perplexity10.1 across 103 answers031% (32 of 103)
Gemini9.2 across 110 answers1120% (22 of 110)
Google AI Mode3.4 across 111 answers2140% (44 of 111)

Read that table as a budget. Google AI Mode cites few sources and leans heavily on Reddit, so a tidy FAQ block on your own domain is competing for very few slots. Perplexity cites a lot per answer, so there is more room, but the room is mostly filled by third-party pages.

How to check it worked: for each of your top five questions, you should be able to name the three or four domains that keep showing up in the answer. If your domain is not among them, the FAQ block is a nice-to-have, not the fix.

Common mistake: assuming schema is what gets you cited. The engines cite pages; schema just makes a page easier to parse once it is already in the running.

4. Decide between FAQPage, HowTo, and other structured data for the same question

Not every question belongs in an FAQ block. A question like "how do I connect Stripe" is a procedure, and marking it as a one-paragraph FAQ answer flattens it.

FormatBest forRich result still shown?Risk of being ignored
FAQPageShort factual Q&A: pricing rules, limits, eligibility, support scopeMainly for government and health sitesHigh for most SaaS sites, though the markup is still parsed
HowToStep-by-step procedures with an ordered sequenceNot statedModerate; treat the visible steps as the deliverable
Product / Offer / SoftwareApplicationPlan names, prices, ratings on a pricing or product pageNot statedLower, because it describes an entity rather than a layout
No schema, clean HTML headingsLong, nuanced answers that need subheads and examplesNot applicableNone; crawlers still read H2s and paragraphs

The practical rule: if the answer fits in two or three sentences and has a definite end, FAQPage. If it has numbered steps, HowTo plus visible numbered steps. If the explanation runs long, write it as a proper section with an H2 that is the question. That last option is underrated. An H2 phrased as the buyer's question, followed by a direct answer in the first sentence, is the most liftable thing on a page, with or without markup.

Common mistake: stacking FAQPage onto a page that is already an article with Article schema and a HowTo block, hoping one of them sticks. Pick the one that describes the page.

5. Write the markup so the answer text matches the visible page exactly

The rule Google states plainly: the question and answer must both be visible to the user on the page. Hidden FAQ content, or answers that exist only in the JSON-LD, is a violation.

Do it in this order:

  1. Write the visible FAQ section first. Question as an H3, answer as plain paragraphs underneath.
  2. Copy that text into the JSON-LD. Literally copy it.
  3. Keep the acceptedAnswer text as simple HTML. Links and basic formatting are allowed inside the answer; scripts and widgets are not.
  4. Put one FAQPage block per page. If the same three questions appear on eight pages, mark them up on the one page that is actually about them.

How to check it worked: view source on the live page, find the JSON-LD, and diff the answer strings against the rendered text. If your FAQ accordion is loaded by JavaScript after the fact, check the rendered DOM in Google's URL Inspection tool, not just the raw HTML.

Reading this because you need customers?

Paste your URL. See what ChatGPT, Perplexity and Google AI answer when your buyers ask, who they name instead of you, and which pages they read. Then we do the work to get you onto those pages.

Free scan, no signup, no card.

Common mistake: writing a longer, keyword-stuffed answer in the schema than the one on the page, because "it is for the crawler anyway." That is the fastest way to get the whole block ignored.

6. Test the markup and know when Google will ignore or restrict it

Three checks, in this order:

  • Schema.org validator, for syntax. Catches broken JSON and wrong property names.
  • Google's Rich Results Test, for eligibility. It will tell you the FAQ block is detected even when no rich result is being shown for your site type.
  • Search Console's Enhancements reports, for whether Google keeps seeing the markup over time on live URLs.

Expect "valid but not showing." That is the normal outcome for a SaaS site now, and it is not an error to chase.

Google will ignore or flag FAQ markup when the content is not visible, when the questions are promotional rather than informational, when the same block is duplicated across the whole site, or when the answers are thin filler written to hold keywords. Manual action for structured data spam is real, and it costs more than the markup ever returned.

Common mistake: re-submitting the same page for indexing every week because no rich result appeared. The rich result is gated by site type, not by your patience.

7. Do not expect FAQ schema to help with voice search or smart speakers

Do not expect FAQPage markup to get you a placement in voice assistant or smart speaker answers; nothing supports that outcome today.

Treat voice as a side effect of being the clearest answer on the web, not as a channel you can configure. If your answer is short, direct and matches the question wording, it has a chance anywhere. If it needs a rich result to be found, it is already too fragile.

Common mistake: building a large FAQ page specifically "for voice." You have built a page for nobody.

8. Track whether buyers actually see your answer in AI tools, not just Google

Once the markup is live, the measurement question changes. Rankings tell you about a results page. They tell you nothing about what ChatGPT said when someone asked which tool to use.

And the answers move. In welaunch's test of live AI answers (27 Sep 2026, 06-aeo-live.md, variance section), only 9 of 32 brand appearances in ChatGPT answers (28%) were named in all four runs of the same prompt, and 6 of 21 in Perplexity (29%). The brand sets ChatGPT named for the same prompt across four runs overlapped by a mean Jaccard similarity of 0.54 over seven prompts that named any brand, and Perplexity by 0.57 over five prompts, per the same test. Checking once a month tells you almost nothing.

This is the gap welaunch fills: it asks your buyers' questions on ChatGPT, Perplexity, Gemini and Google AI several times a week, keeps every answer in full so you can read what each engine actually said, and finds the pages the engines cite for those questions.

How to check it worked: you should be able to say, for each of your five priority questions, how often you were named over the last month and which domains kept getting cited instead.

Common mistake: screenshotting one good ChatGPT answer and calling it a result.

9. If your FAQ answers never get cited, get them onto the pages AI engines actually pull from

When you have run the tracking for a few weeks and your own FAQ page never appears, the honest conclusion is that the engines are answering from somewhere else. Usually community threads and listings.

In welaunch's test (27 Sep 2026, 06-aeo-live.md), 106 of 434 AI answers (24%) cited reddit.com. Broken out by engine in the same test: Google AI Mode 40% (44 of 111), Perplexity 31% (32 of 103), Gemini 20% (22 of 110), ChatGPT 6% (7 of 110). If Google AI Mode is where your buyers are, a Reddit thread may matter more than your own FAQ block.

That is the second half of what welaunch does: it drafts the reply, listing or page that gets you onto a cited source, works from your own accounts, and leaves the submit button to you. It never posts under an account of its own, and on Reddit and Quora it names the product only, with disclosure, rather than stuffing in links.

Common mistake: writing another FAQ page when the answer the engine is reading lives on a forum you have never posted in.

FAQ

What is a FAQ schema?

FAQ schema is structured data, normally JSON-LD, that labels a list of questions and answers on a page using the FAQPage type from Schema.org. Each entry is a Question with an acceptedAnswer. It does not add content. It describes content that is already visible to visitors.

What is FAQ schema markup used for on a SaaS site?

Mostly for short, factual buyer questions on pricing, security, integration and support pages. It makes those answers easy for crawlers and AI systems to parse as discrete Q&A pairs. It is not a good fit for long explanations, which belong in normal sections with question-shaped H2s.

Is FAQ schema still relevant in 2026?

It is still a valid and parsed type, but do not expect a rich result for most SaaS sites; use it because it gives search engines and AI crawlers a clean, parseable Q&A block. Use it where the content genuinely is a Q&A list. Do not add it expecting a bigger search listing.

Is FAQPage schema deprecated by Google?

No. FAQPage schema is not deprecated; Google still documents it and still reads it. What was cut back is the rich result it used to trigger. Check Google's Search Central updates log if you want to see whether that policy moves again.

Can you provide an example of FAQ schema?

A minimal FAQPage block looks like this, placed in the page head:

{
 "@context": "https://schema.org",
 "@type": "FAQPage",
 "mainEntity": [{
 "@type": "Question",
 "name": "Do you offer a free trial?",
 "acceptedAnswer": {
 "@type": "Answer",
 "text": "Yes. Every plan includes a 14-day trial and no card is required to start."
 }
 }]
}

Replace the question and answer with the exact text shown on your page. The strings must match what a visitor reads.

What are the current ranking factors for getting a featured answer from FAQPage markup?

There is no separate ranking system for FAQPage markup. Eligibility for FAQ rich results is restricted by site type, and beyond that the usual signals apply: the page has to be indexed, relevant to the query, and the answer has to be on the page in readable text. Marking up a weak answer does not promote it.

Sources

Related on welaunch

Do buyers find you, or your competitor?

Paste your URL. See what ChatGPT, Perplexity and Google AI answer when your buyers ask, who they name instead of you, and which pages they read. Then we do the work to get you onto those pages.

Free scan, no signup, no card.

Part of Technical AEO, in the guides on getting your SaaS named in AI answers.

FAQ schema for SaaS: when FAQPage markup helps AI answers and…