FAQ schema markup in 2026: what Google actually surfaces

In August 2023, Google announced it was scaling back FAQ rich results. The accordion-style "People also ask" expansion that used to push your competitors below the fold disappeared from most SERPs almost overnight, and a wave of articles declared FAQ schema dead.

Two years on, the picture is more interesting than "dead" or "alive." FAQ schema is still worth implementing in 2026 — just for different reasons than in 2022. Here's the actual landscape.

What changed in August 2023

Google's announcement (paraphrased): FAQ rich results would only show for "well-known, authoritative government and health websites." For everyone else, the rich result vanished.

What actually happened:

• The accordion rich result disappeared from ~95% of queries it used to show on, including most commercial and informational pages.
• The schema itself kept being parsed — Google didn't ignore it, just stopped rendering the rich result.
• Branded queries and high-authority sites kept some FAQ rich results, often inconsistently.

Most SEO platforms removed FAQ schema from their default templates. Many brands ripped it off their pages. Some did it permanently, which is the move I want to argue against.

What FAQPage schema still does in 2026

Even without the SERP rich result, your FAQ schema is still working in four places:

1. AI Overviews and generative search citations

This is the big one. Google's AI Overview, ChatGPT search, Perplexity, and Bing Copilot all consume structured data when picking sources to cite. Pages with clear question/answer structure (whether via FAQPage schema or natural Q&A formatting) cite at noticeably higher rates than pages with the same content in flowing prose.

The schema isn't a magic switch — your content still has to be authoritative — but it gives the model a parseable, low-ambiguity signal: "here is question X, here is the canonical answer Y." Models love that.

2. Voice search and assistants

Google Assistant, Siri, and Alexa all draw heavily on schema-marked content for answer cards. FAQPage schema is one of the cleanest paths into a voice answer for question-shaped queries.

3. Bing rich results

Bing didn't follow Google's deprecation. FAQ rich results still show on Bing for general-purpose pages, and Bing's market share grew alongside its Copilot integration.

4. Internal Site Search and CMS surfaces

Your own site search (Algolia, Typesense, Pagefind, etc.) can use FAQ schema to surface specific Q/A pairs in instant results. Documentation sites particularly benefit — a help-centre query like "how do I reset my password" can match a single FAQ entry instead of returning the whole article.

What schema does still earn rich results in 2026

For balance, here's what's reliably surfacing in the Google SERP today:

HowTo got deprecated harder than FAQPage and didn't come back. Stop implementing it.

Implementation: the things people get wrong

FAQ schema must match visible content

The single biggest violation: marking up Q/A pairs in JSON-LD that don't appear visually on the page. Google's docs are explicit — every question and answer in your FAQPage schema must be visible on the page. Hidden FAQ markup is a manual-action risk and generally a bad signal.

One FAQPage per page, not per question

Some templates emit a FAQPage element per question. Wrong. The right shape:

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    { "@type": "Question", "name": "...", "acceptedAnswer": { "@type": "Answer", "text": "..." } },
    { "@type": "Question", "name": "...", "acceptedAnswer": { "@type": "Answer", "text": "..." } }
  ]
}

One FAQPage, with mainEntity as an array of Question objects.

Plain text in answers, not HTML

The text field in acceptedAnswer should be plain text. Some teams paste in HTML markup, which validates but renders as escaped HTML strings in voice and AI contexts. Strip the tags before serialising.

Don't include "FAQ" content that isn't actually a question

I see this on landing pages: "What makes us different?" with an answer that's just a marketing pitch. Search engines and language models recognise the pattern. They don't reward marketing copy disguised as Q&A — and increasingly penalise it as low-trust.

Real questions users ask. Real answers. Otherwise skip the schema.

A pre-publish checklist

Before you ship a page with FAQ content:

1. Real questions. Drawn from People also ask, your support inbox, or competitor analysis — not invented marketing prompts.
2. Visible on the page. Every Q/A in JSON-LD must appear on the rendered page.
3. Plain text in acceptedAnswer.text. No HTML.
4. One FAQPage element per page. Multiple questions go inside mainEntity.
5. Validate. Run your JSON-LD through Google's Rich Results Test and Schema.org's validator.
6. Direct, citable answers. Each answer should stand alone if extracted by an AI model — first sentence answers the question, rest provides nuance.

When to skip FAQ schema entirely

Some pages don't need it:

• Pure listicles where each item is a feature, not a Q&A.
• Landing pages with marketing copy disguised as Q&A — these now hurt more than help.
• Pages whose Q&A content is paywalled or hidden behind interactive states.

Tools

• Schema Generator — generates valid JSON-LD for FAQPage, Article, Product, Breadcrumb and Local Business with one click. Copy straight into your <head>.

The TL;DR: FAQ schema lost its SERP rich result but gained a quieter, more durable role in AI citations and voice. If your pages have real Q&A content, mark it up. If they don't, don't fake it.