If you do one piece of structured-data work for AEO this quarter, do FAQPage schema.
It's the lowest-effort, highest-leverage move available. It hands ChatGPT, Perplexity, and Google's AI Overview the structured question-and-answer format they extract and cite first, and it forces you to write direct-answer content that converts visitors better than long marketing prose. (Google retired the old FAQ rich-result snippet in May 2026, so the payoff is AI citation and clarity, not a SERP accordion.)
What FAQPage schema actually is
FAQPage schema is a piece of JSON-LD (JavaScript Object Notation for Linked Data) that you embed in the HTML head of any page with a frequently-asked-questions section. It tells Google, Bing, and AI assistants: "this section is structured Q&A, here's the question, here's the answer."
That structure is exactly what AI extraction is looking for.
The copy-paste core
The exact format
Here's sample JSON-LD for an HVAC company (swap in your own details). It goes in your page's <head>:
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "How much does an HVAC repair visit cost?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Repair visits range from $[low] to $[high] depending on the issue. We give a quote before any work begins."
}
},
{
"@type": "Question",
"name": "Which cities do you serve?",
"acceptedAnswer": {
"@type": "Answer",
"text": "We serve [City A], [City B], and [City C]."
}
}
]
}That goes inside a <script type="application/ld+json"> tag in the page <head>. Each question gets its own object inside mainEntity. Use as many as you need, five to seven is the typical sweet spot.
What makes the schema actually work
Match the visible content. The questions and answers in your schema must match the questions and answers visible on the page. Google penalizes schema that describes content not actually on the page.
Use real customer questions. Not made-up SEO questions. Real ones — the things people actually ask before calling. If you don't know what those are, look at your contact form submissions, your call logs, your competitor FAQ pages.
Write direct answers in plain language. First sentence answers the question. Subsequent sentences add detail. Don't pad. Don't hedge. AI extraction grabs the first 1–3 sentences usually.
Include specifics. Prices, time ranges, locations, real client examples. "We serve Kent, Auburn, and Renton" is much more extractable than "We serve the surrounding area."
How to test it
Two free tools:
- Google's Rich Results Test at search.google.com/test/rich-results, paste your URL, get a structured-data validation report. Your FAQPage schema should validate cleanly; that's what matters for AI extraction now that Google has retired the FAQ rich result.
- Schema Markup Validator at validator.schema.org, more general schema validation, good for catching syntax errors.
Both should return zero errors before you publish.
Real example
Pinnacle Air Control's South Hill HVAC site uses FAQPage schema on the homepage and on each service page. The site was ChatGPT's #1 recommendation for "heating repair south hill" in 2026, and its FAQ answers are the exact chunks AI assistants extract. Both outcomes are downstream of the FAQ schema doing its job.
The implementation took maybe two hours of focused work, including writing the answers. The ranking improvements compound monthly.
What FAQPage schema isn't
FAQPage schema isn't a hack or a shortcut. Google explicitly supports the markup, AI assistants explicitly extract from it, and the work itself improves the page for human readers. There's no penalty risk if you follow the rule: match visible content, use real questions, write real answers.
FAQPage schema is one of the structured-data items I add as part of SEO & AEO optimization when I take on a client's ongoing AEO work.