Not in Google, not anymore. Google stopped showing FAQ rich results on May 7, 2026, after limiting them to government and health sites in 2023. The markup no longer earns you anything on the results page. The questions themselves, written on the page, still do the work.
Does FAQ schema still work in 2026?
If "work" means the expandable questions under your result in Google, no. That feature is gone. Google announced it would stop appearing on May 7, 2026, and in June it removed the documentation for it, noting that the FAQ rich result is no longer shown in Google Search.
If "work" means helping a page explain itself, the answer is less dramatic. A page that lists the real questions people ask and answers each one in a sentence or two is easier to read for a person and easier to quote for anything that reads it. That part never depended on the markup.
What did Google change about FAQ rich results?
It happened in two steps, both in Google's own changelog for Search Central.
| When | What Google changed | What it meant for a small site |
|---|---|---|
| September 2023 | FAQ rich results shown only for well-known, authoritative government and health websites | Almost every business lost them overnight |
| May 7, 2026 | FAQ rich results stop appearing in Google Search | Nobody gets them, whatever the markup says |
| June 2026 | Documentation for the FAQ rich result removed | No official guide left to follow |
The same changelog shows the wider pattern. Since 2025 Google has been retiring several structured data features to simplify the results page. FAQ was one of the most used, so it was one of the most noticed.
Is FAQ markup useful if no rich result shows?
Less than it used to be, and I won't pretend otherwise. I can't show you that FAQ markup makes Google or an AI assistant cite a page more often, because nobody outside those companies can measure that cleanly. What I can say is narrower: the markup describes, in a format machines read without guessing, which questions the page answers and what the answers are.
I still generate it on every article here, from the questions visible on the page, with a small script that refuses to mark up anything a reader can't see. It costs nothing once it's automatic. If yours takes hours to maintain by hand, the hours are better spent on the questions themselves.
The one rule that matters either way: never mark up questions that aren't on the page. Structured data that says something the visible page doesn't is the kind Google treats as a problem.
How many service sites still mark up their questions?
Fewer than you might expect, and the gap is between having questions and marking them. In October I re-read the same 54 US therapy practice websites from my September sample, the same 418 pages, without JavaScript, the way a crawler gets them. Three sites didn't answer this time.
- 13 of the 54 sites served FAQPage markup on at least one page, 20 pages in total.
- 21 sites had a page of frequently asked questions among the pages I read. Only 7 of them marked it up.
- Of the 24 FAQ pages I could read, 6 carried FAQPage markup.
- 49 of the 54 served some kind of structured data, usually about the website or the business, not about the questions.
So most practices that wrote an FAQ page never told a machine what was on it, and since May that costs them nothing in Google. The question worth asking now isn't whether to add the markup. It's whether the questions on the page are the right ones, and whether the answers are short enough to quote.
How do you write questions worth answering on the page?
Start from what people actually ask, not from what you'd like them to ask. The best list comes from your inbox and your first calls: the price question, the "do you see people like me" question, the "what happens in the first session" question.
Then write each one the way it would be typed. "How much does a session cost?" beats "Our fees". Answer it in the first sentence, add the detail after, and keep each answer to a few lines. If an answer needs a page, it deserves its own page, linked from the question.
And put the questions where the decision happens. An FAQ page nobody visits answers nobody. The same three questions under the service they belong to get read. I wrote about structuring pages so a machine can carry them away in how to make a website citable in AI search, and about pages that can be compared and chosen in AI agents and the boutique website.
Questions I get about this
Should I remove my FAQ schema now?
There's no reason to rush. It no longer produces a rich result in Google, but if it matches what's on the page it describes the page correctly. Remove it if it's out of date or covers questions you deleted.
Does FAQ schema help with ChatGPT or other AI assistants?
I don't know, and I haven't seen anyone show it with data they let others check. Clear questions and short answers on the page are the part you can control and see.
Is an FAQ page still good for SEO?
An FAQ page is as good as its questions. Real questions with direct answers help people decide; a page of filler questions helps nobody, with or without markup.
What about llms.txt, does it replace structured data?
No, they do different jobs. I covered what belongs in one in what to put in an llms.txt file.
What I can and can't claim
The changes to FAQ rich results come from Google's Search Central documentation and its changelog, checked on October 7, 2026. The numbers come from 54 private therapy and counseling practice websites in ten US metros, the same sample and pages as my September measurement, re-read on October 7, 2026 without JavaScript; 392 of 418 pages answered. It's a convenience sample of one kind of business. What I can't tell you is whether FAQ markup changes how often an AI assistant cites a page. Nobody outside those companies can measure that cleanly, so I don't claim it.
If you want me to look at which questions your pages answer and which ones your clients actually ask, that's part of the audit, or you can book a strategic fit call.
How this was made: the same 418 pages from 54 public therapy practice websites, re-read by Contento on 7 October 2026 (392 answered), with the method described above. Drafted with AI assistance, then checked against the data, edited and signed by Lucas Fernández Díaz.
