On 7 May 2026, Google quietly closed a chapter that most SEO teams had been writing for six years. FAQ rich results — the expandable Q&A boxes that used to dominate the search results page — were officially retired. The structured data report and the rich results test will both drop FAQPage support in June. For a lot of marketers, the first instinct was to strip the markup out of their templates.

That would be a mistake. While Google was deprecating the visual feature, the same FAQPage schema quietly became one of the most consistently cited content formats inside ChatGPT, Perplexity and Google's own AI Overviews. Recent analysis suggests pages with FAQPage schema are roughly 3.2x more likely to appear in AI Overviews, and one study found a 28% lift in citation rates versus pages without it. AI-surfaced URLs also skew about 25.7% fresher than traditional search results, based on a sweep of 17 million citations.

This is a guide to structuring FAQ pages so AI engines quote you. We'll cover what FAQ schema actually does in 2026, why answer engines lean on this format, how to write a Q&A that gets extracted verbatim, what the markup should look like, and how to measure whether it's working.

What is FAQ schema and is it still worth using in 2026?

FAQ schema is a JSON-LD markup standard (FAQPage from schema.org) that labels each question and its answer on a page so search engines and AI systems can parse the Q&A structure without guessing. It is still worth using in 2026 — Google has dropped the rich result, not the standard, and pages that keep valid FAQPage markup are not penalised in any way.

The reason it matters more, not less, is who's reading it now. The Google blue link audience used to be the audience that benefited. In 2026 the bigger reader is the answer engine — ChatGPT search, Perplexity, Gemini, Claude — looking for clean, self-contained Q&A pairs it can cite. At Vridhii, every client site we ship still includes FAQPage schema on the pages that are designed to be quoted. The cost is one block of JSON; the upside is a 28% better chance of being the source the AI names.

There is a real debate worth flagging. A controlled experiment in February 2026 showed that large language models tokenise JSON-LD as raw text, so the schema itself doesn't magically teach the model semantics. The visible Q&A on the page does most of the heavy lifting. But the schema still helps indirectly: it improves how Google's Knowledge Graph indexes the content, and Google AI Overviews leans on that index. Belt and braces, in other words. Keep the schema; just don't rely on it alone.

Why do AI engines quote FAQ pages more than other formats?

AI engines quote FAQ pages more than other formats because Q&A pairs are the smallest, most self-contained unit of meaning on the web — one question, one direct answer, no surrounding context required. That's exactly the shape an answer engine needs when it's asked a question and has to return a sentence or two with a source.

Three things stack up in favour of the format. First, the question itself is a near-perfect match for how people prompt — "what is", "how do I", "does it work" — so the AI's retrieval layer pulls the page in early. Second, the answer is bounded: an FAQ trains the writer to be direct in the first sentence, which is the sentence the AI quotes. Third, the structure is repeatable. A page with eight clean FAQs gives the model eight chances to cite you. A 1,500-word essay gives it one tangled paragraph to summarise.

"In 2026, the unit of content that gets cited isn't the article. It's the answer."

How should you structure each question and answer for citation?

Structure each FAQ as a real question phrased in your customer's words, followed by a self-contained answer that resolves the question in the first sentence and runs 40–60 words. The single biggest predictor of citation is whether the first sentence of the answer can be lifted out of the page and still make sense.

A few rules we apply to every Vridhii client FAQ:

What does the right FAQ page look like in HTML and JSON-LD?

A well-structured FAQ page in 2026 has three layers: visible Q&A with semantic headings, a matching FAQPage JSON-LD block, and a credible byline plus a last-updated date. The visible content is what the AI reads. The JSON-LD reinforces it. The byline and date make the page eligible to be trusted.

The pattern below is what we ship by default. It works whether the FAQ lives on a service page, a blog post or its own dedicated URL:

<section itemscope itemtype="https://schema.org/FAQPage">
  <div itemscope itemprop="mainEntity" itemtype="https://schema.org/Question">
    <h2 itemprop="name">How much does FAQ schema affect AI citations?</h2>
    <div itemscope itemprop="acceptedAnswer" itemtype="https://schema.org/Answer">
      <p itemprop="text">FAQ schema indirectly improves
        AI citation rates by about 28%, mostly because the visible
        Q&A format is easy for answer engines to extract...</p>
    </div>
  </div>
</section>

Add the JSON-LD version of the same Q&A in a <script type="application/ld+json"> block at the bottom of the page. The two should match word for word — drift between visible content and schema is the most common technical error we see when we audit client sites. Google's documentation explicitly says unused FAQPage markup will not hurt rankings, so you can keep older FAQ blocks in place without worry while you refactor.

How do you measure whether your FAQ pages are getting cited?

Measure FAQ citation performance on two tracks: traditional impressions in Google Search Console, and direct AI citation tracking through tools that scrape ChatGPT, Perplexity and Gemini for brand mentions. The traditional metrics tell you whether the page is being indexed and surfaced; the AI metrics tell you whether the page is being quoted.

A practical monthly review for a small marketing team looks like this. In Search Console, filter for queries phrased as questions and watch impressions and average position trend on your FAQ URLs. In a tool like Profound, Otterly or Goodie, set up brand and topic monitors for the questions your FAQs answer, and log how often your domain shows up in AI answers. And once a quarter, run the same prompts manually across ChatGPT, Perplexity and Gemini and screenshot the sources. If your URL is in the citation pile, the FAQ is doing its job. If it isn't, the question phrasing or the answer's first sentence is usually where to start rewriting.

Key takeaway: FAQ schema for AI is not dead — Google retired the rich result, not the standard. Pages with FAQPage markup are roughly 3.2x more likely to appear in AI Overviews and see about 28% higher citation rates across answer engines. The format wins because Q&A is the smallest, most extractable unit of meaning on the web. Structure questions in the user's words as H2s, answer in 40–60 self-contained words with the answer in the first sentence, mirror it in JSON-LD, and keep the page fresh.

Frequently Asked Questions

Frequently asked questions

Did Google ban FAQ schema in May 2026?

No. Google retired the FAQ rich result — the expandable Q&A box in search — on 7 May 2026, but FAQPage schema itself is not deprecated. Google has confirmed that keeping valid FAQPage markup will not hurt rankings, and the schema continues to feed AI Overviews and the Knowledge Graph.

How long should each FAQ answer be for AI citation?

The sweet spot is 40–60 words, with the direct answer in the very first sentence. That length is long enough to add useful context and short enough that an answer engine can quote the whole thing without truncating. Anything beyond about 80 words is usually paraphrased instead of cited.

Where should FAQs sit on the site — one big page or scattered across content?

Both, but with intent. Keep a dedicated FAQ hub for evergreen questions about your company and pricing, and embed 3–6 highly relevant FAQs at the bottom of every service page and blog post. Scattered, contextual FAQs tend to earn more AI citations because they sit next to the topic the user is researching.