If you ship one schema change on your site this quarter, ship FAQPage. The correlation between FAQPage schema and AI citation is the strongest single signal we see across audited sites in 2026. Pages with FAQPage schema are roughly 2 to 3 times more likely to appear in a ChatGPT or Perplexity answer than pages with only the WordPress and Yoast defaults. And yet, in audits of 100 mid-market B2B sites, fewer than 1 in 4 had FAQPage schema deployed on any commercial page.
This article covers why it works, where to put it, the actual JSON-LD you can paste, and the three failure modes that make most FAQPage implementations not count.
FAQPage is a schema.org type that wraps a list of question-and-answer pairs in machine-readable JSON-LD. When an AI search engine crawls your page, the schema gives it a clean, pre-structured set of facts it can lift directly into a generated answer. The engine does not have to infer that a paragraph is the answer to a question. You hand it the question and you hand it the answer.
That sounds basic. The reason it matters is that AI engines are not just extracting from your page content. They are scoring how confidently they can attribute a clean answer to a source. FAQPage schema is a confidence signal. It says: this site told me, in structured form, that this answer goes with this question.
Three reasons, in roughly the order they cost sites citations:
1. The marketing team writes FAQ sections without wrapping them in schema.
The visible HTML on the page has the questions and answers. The JSON-LD does not. Yoast does not auto-detect FAQ sections in arbitrary content blocks. Unless someone manually deploys the schema, the page ships with all the citation-friendly content and none of the citation-friendly markup. We see this on roughly 60 percent of the audited sites where FAQ sections are visible on the page.
2. The CMS has a FAQ block but the schema is malformed.
Block-based editors (Gutenberg in particular) sometimes generate FAQ blocks that emit JSON-LD with missing required fields, broken @id references, or duplicate Question entries. Google’s Rich Results Test will often pass these. Real AI engines are stricter. We see this on roughly 15 percent of the audited sites that have FAQPage in their HTML.
3. The FAQ questions are written in a way no buyer would actually ask them.
Schema is plumbing. The questions are the actual signal. Pages that wrap obviously-promotional questions like “Why is [Brand] the best at X?” in FAQPage rarely get cited, because no buyer asks AI engines a question phrased as a brand pitch. The questions need to match what buyers type. We see this on roughly 25 percent of audited sites with FAQ schema.
The compound effect is severe. Of the 25 percent of sites that ship FAQPage at all, maybe half ship it well. The remaining 12 percent are getting the citation lift. The other 88 percent are leaving it on the table.
Not every page. FAQPage on every page dilutes the signal and looks structurally desperate. The right pages are:
For NetPageTwo’s own site, that worked out to six pages. The site ships FAQPage on the homepage, /the-new-search/, /pricing/, /geo-101/, /vs/otterly/, /vs/profound/, and /vs/seo-agency/. The rest of the site does not need it.
If you have product pages, the same rule applies: only the product pages where buyers actually ask questions get FAQPage. A category page does not.
Here is a working FAQPage block. Paste it before the closing body tag or inside any HTML block in your CMS. WordPress users: use a Custom HTML block or a Code Snippets plugin.
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "FAQPage",
"@id": "https://yourdomain.com/page-slug/#faq",
"mainEntity": [
{
"@type": "Question",
"name": "How does your service handle [the thing buyers ask about most]?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Three or four sentences of a real, specific answer. Not a tagline. Numbers and named processes if they apply. The text here is what AI engines will lift."
}
},
{
"@type": "Question",
"name": "What is the second most common question?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Same pattern. Real answer, specific, no marketing language."
}
}
]
}
</script>
Two notes on this block.
The @id field is optional but recommended. It gives the FAQ a stable identifier the rest of your schema graph can reference. If you also ship Organization or Service schema, you can use the @id to link them.
The name field of each Question is the literal text that AI engines treat as the query. Phrase it like a buyer would. Not “What makes us different?” but “What is the difference between productized SEO and a traditional agency?”. The first is a brand pitch. The second is a query.
Failure mode 1: leaving the JSON-LD broken.
Always validate. Run https://search.google.com/test/rich-results on the page after pasting. If it returns errors, fix them before moving on. Most errors come from missing required fields (acceptedAnswer is mandatory), duplicate @id values across the page, or unescaped quotation marks inside the answer text.
Failure mode 2: making the questions promotional.
Read the questions out loud. If they sound like the script of a sales call, rewrite. The questions should sound like what someone would type into ChatGPT or Google search. Often the question that buyers actually ask is the one your sales team has been answering on every fit call for a year.
Failure mode 3: making the answers too long.
AI engines extract answers most reliably when they are 40 to 120 words. Shorter is fine. Longer makes extraction worse, because the engine has to decide which sentence is the actual answer. Three sharp sentences beat a paragraph.
Imagine you run a productized accounting service for ecommerce brands at $349 a month. Your most common pre-sale question is “Is this just bookkeeping or do you also handle the tax filings?” Wrap that as a Question. The accepted answer:
“Both. The $349 tier covers monthly bookkeeping, quarterly P&L review, and federal plus state tax filings for one entity. If you operate multiple legal entities or need international filings, talk to us about the $649 tier.”
Three sentences. Specific. Numbers. A real boundary. AI engines will lift this verbatim when someone asks ChatGPT “is X an accounting service or just bookkeeping?”.
The pattern is: a real question buyers ask, a real answer you’d give on the phone, structured in JSON-LD so the engine can extract it cleanly. Everything else is plumbing.
Does Google still show rich results for FAQPage?
Not since 2023. Google restricted FAQ rich results to authoritative health and government sites. But the schema still works as a citation signal for AI engines, which is the larger use case in 2026. You are shipping schema for ChatGPT and Perplexity, not for the Google SERP feature.
Will deploying FAQPage hurt my SEO?
No, unless you stuff the page with 30 FAQs you invented to sneak keywords. A focused 4-7 question FAQ that answers real buyer queries is a positive signal across both classical SEO and AI search.
How long until I see citation lift?
Bing reindexes most sites within 24-72 hours of a sitemap ping, and ChatGPT browse uses Bing-routed retrieval. So technically: days. Realistically: 14-30 days before you can attribute citation gains to the change, because traffic data is noisy at small scale.
Do I need a plugin?
No. A Custom HTML block with the JSON-LD pasted in works on any WordPress site. The plugins (Yoast, Rank Math, Schema Pro) automate the deployment but the raw schema is what counts. If you can paste, you can ship.
If you have not deployed FAQPage on your top 5 pages by next Friday, you are leaving the highest-correlation citation signal on the table. The deployment time per page is 30-60 minutes including validation. The longest part of the work is writing the right 4-7 questions.
If you want us to do it on your top 10 pages this week, book the fit call. We bring the audit to the call, you see which pages get FAQ first, and the schema ships in week one. Productized SEO and AI search at $489 a month, month to month, cancel in 14 days.
Related Field Notes:
– Schema markup for AI search: a 2026 field guide
– Answer capsules: the first 60 words that matter for AI extraction
– GEO 101: Generative Engine Optimization Explained
– Net Page Two vs Otterly: tracking tool vs productized service