If you are trying to figure out whether FAQ schema is still worth having now that Google has retired FAQ rich results, here is the direct answer. The expandable question and answer boxes are gone from search results for everyone, but the markup itself still works, Google still parses it, and removing it gains you nothing. What got deprecated was a display feature. What FAQ markup contributes to AI understanding was never the display.
The timeline matters because it explains the confusion. In August 2023, Google restricted FAQ rich results to a small set of government and health websites, which made them invisible for nearly every business. In May 2026, Google retired the feature entirely, and the reporting for it in Search Console followed. So the visible reward that motivated most FAQ schema projects has been dying in stages for three years, and it is now fully dead.
That is why “should I remove FAQ schema” has become a common question. It is usually shorthand for a bigger one: does any of this still matter now that AI systems answer questions directly? That bigger question has a real answer, and it starts with understanding what the rich result actually was.
The Rich Result Was Never the Part AI Used
A rich result is a presentation decision. Google reads your markup and chooses to decorate your listing with it. Comprehension is a separate layer. Google has said that it continues to parse FAQ markup to understand pages and that leaving valid markup in place causes no problems. The decoration ended. The reading did not.
The reason this feels confusing is that for years, businesses judged schema by what it changed on the screen. FAQ schema was popular precisely because it produced something you could see: your listing took up more space, your questions appeared under your link, and the effect was easy to show a client or a boss. When the visible effect disappeared, the natural conclusion was that the markup stopped working.
But that conclusion mixes up two different jobs. The screen effect was marketing real estate. The markup’s other job, telling machines exactly which text on the page answers which question, kept right on functioning. This is the same distinction drawn in Schema Does Not Change What AI Thinks. It Changes What AI Knows: schema affects what systems can extract with confidence, not how your page looks or how persuasive it is. The extraction job was always the durable one. The decoration was always the temporary one. It just took Google retiring the feature to make that obvious.
What FAQ Markup Actually Does for AI
FAQPage markup pairs each question on a page with the complete text of its answer. That is the entire function. It draws a clean boundary around each answer so a machine does not have to guess where one ends and the next begins, and it states plainly that this text exists to answer that question.
For AI systems, that structure lines up with how they work. When ChatGPT, Claude, Gemini, or an AI search feature answers a question, it favors passages that stand on their own: a clear question followed by a direct, complete answer that makes sense without the rest of the page. A genuine FAQ is already in that shape. The markup makes the shape explicit.
Here is the honest caveat, and it is the same one that applies to every schema type. The markup labels the content. It does not improve the content. As covered in What Schema Actually Does, schema cannot add a claim that was never written. AI systems pull from clear question and answer content whether or not the markup is present, so the active ingredient is the answer itself. FAQ markup on a vague answer produces a clearly labeled vague answer. If you have to choose where to spend effort, the writing wins every time. The markup is worth keeping because it costs nothing to keep and removes guesswork for the machines that still read it. It is not worth treating as a strategy on its own.
The Part of Your FAQ That Matters More Than Any Markup
There is a reason FAQ content keeps showing up in AI answers, and it has little to do with schema. On most business websites, the FAQ page is the only place the business answers questions the way a customer actually asks them. What exactly do you do. Who is this for. What does it cost. How long does it take. What happens if it does not work.
Everywhere else on the site, those questions get answered in marketing voice, if they get answered at all. The homepage gestures at outcomes. The services page describes capabilities. The FAQ page, almost by accident, is where the plain statements live. And plain statements are what AI systems can quote, because a direct answer survives being lifted out of context and a slogan does not. That is the point made in Why Clear Writing Beats Clever Writing for AI, and an FAQ is often the clearest writing a business publishes.
Which means the worst possible response to the rich result going away is the one some businesses are choosing: deleting FAQ pages because the feature that rewarded them is gone. That removes some of the most extractable, most quotable content on the site to escape a penalty that does not exist.
Where FAQ Content Quietly Goes Wrong
There are two failure modes worth checking for, and neither one shows up in a schema validator.
The first is fabricated FAQs. When FAQ schema was producing rich results, plenty of sites invented questions nobody asks so they could occupy more of the search page. That incentive is gone, but the pages remain. An FAQ full of manufactured questions with thin answers is not neutral. It is a set of plainly labeled statements about your business that say very little, sitting on exactly the kind of page AI systems read closely.
The second is contradiction. FAQ pages get written once and forgotten, while the rest of the site gets updated. Prices change, services change, service areas change, and the FAQ keeps confidently stating the old answer. Because FAQ answers are direct and quotable, an outdated FAQ is not a minor inconsistency. It is a clear, extraction friendly statement of wrong information, competing against the correct information elsewhere on your site. When AI systems find both, the result is the kind of conflict described in What Happens When AI Finds Conflicting Information About a Business: hedged answers, blended answers, or the wrong fact stated with full confidence.
And there is one structural limit to keep in mind. FAQ markup describes questions and answers on a page. It does not describe your business as an entity. What your company is, what it is known for, and how it relates to its people and services is the job of Organization schema and of the writing on your core pages, the same division of labor covered for blog posts in Article Schema and AI. A thorough FAQ answers customer questions. It does not, on its own, establish identity.
The Question a Validator Cannot Answer
Everything above ends in the same place. Your FAQ markup can be valid, your rich results report can be empty because the feature no longer exists, and none of that tells you the one thing worth knowing: when someone asks ChatGPT, Claude, or Gemini the questions your FAQ was written to answer, what do they actually say?
That is the gap the AI Business Understanding Report is built to close. Instead of checking whether your markup is technically correct, it looks at what the major AI systems currently say about your business, including whether the plain answers you published are the answers coming back, and where old, thin, or conflicting content is shaping the response instead. If your FAQ is doing its job, the report shows that. If it is quietly feeding AI outdated or vague answers, the report shows that too, along with what specifically needs to change.
FAQ rich results are gone, and they are not coming back. The markup is fine to keep. The content underneath it is where the real leverage has been all along. If you want to know whether that content is producing accurate answers about your business right now, order a report and find out what AI actually says, instead of guessing from a feature that no longer exists.