Yes — FAQs matter more for AI search in 2026 than they ever did for SEO, but not as schema and not as a block bolted to the bottom of a page. Google removed FAQ rich results from Search entirely on May 7, 2026, so the expandable dropdown you used to earn under a listing is gone for every site, and there is no confirmed evidence that FAQPage markup by itself wins you a single AI citation. What assistants reward is the shape — a real question with an answer that stands on its own — wherever it sits on the page.
We build an AI-visibility plugin for WordPress, and the pattern we watch play out is that sites keep optimising the wrong half of the FAQ. They chase the markup and the labelled "FAQ" region, both of which are now near-worthless, while ignoring the thing that actually gets lifted into an answer: a self-contained question-and-answer chunk, whether or not anything calls it an FAQ.
Do FAQ blocks still help now that Google removed the rich result?
A dedicated FAQ block only helps if its content is genuinely unique and answers each question in a self-contained way — the SERP feature it used to earn is gone. Google's wind-down ran in stages, and as of mid-2026 there is nothing left to win from the display side.
Aug 2023
FAQ rich results limited to authoritative government and health sites
May 7 2026
FAQ rich results stop appearing in Google Search for every site
Jun 2026
Search Console FAQ report and Rich Results Test FAQ support removed
Aug 2026
Search Console API support for FAQ rich-result data scheduled to end
So the reason to keep an FAQ section in 2026 is no longer the dropdown — it's whether the questions and answers earn their place as content. If your FAQ repeats what the article already said, it earns nothing. If it answers something the body never addressed, in a passage that reads cleanly out of context, it's as liftable as any other part of the page. The questions worth adding first are the ones even top-ranking competitors leave unanswered, because an assistant has no strong source to pull from there. The label "FAQ" adds nothing either way; the standalone answer does the work.
How do AI assistants actually use Q&A content?
They break a page into chunks and score each passage on its own, so a question with a self-contained answer is close to the ideal unit — small, complete, and easy to lift into a reply with a citation. This is why Q&A content punches above its weight in AI answers: retrieval systems don't read your page top to bottom, they hunt for the single passage that best matches a prompt, and a tight question-and-answer is exactly that passage.
The catch is that this has nothing to do with a region labelled "FAQ." An answer-first section anywhere in your body — a question-shaped heading with the answer in the first sentence — produces the same liftable chunk, and usually a better one, because it's where the reader is actually looking. The mechanics of building that chunk, heading plus answer sentence plus the fact underneath, are the whole subject of answer-first writing, and once a page is written that way, a separate FAQ block becomes redundant rather than essential.
Does FAQPage schema still earn you AI citations?
There is no confirmed evidence that it does. Google's own guidance says no special schema is required for AI Overviews or AI Mode, FAQPage remains a valid Schema.org type that Google will still parse to understand a page, and assistants like ChatGPT and Perplexity pull clean Q&A content whether the markup is present or not. The active ingredient is the text, not the tags around it.
That leaves schema as low-cost and low-reward: harmless to keep, not worth new effort, and definitely not the visibility hack some guides still sell it as. If a plugin already emits FAQPage markup for you, there's no need to rip it out — just don't expect it to move anything. The honest version of the 2026 advice is to stop treating markup as the lever and start treating the passage as the lever, which is the same reordering of priorities behind every one of the reasons in why AI assistants never cite your blog.
Why can a dedicated FAQ block actually hurt?
A bolted-on FAQ block hurts when it duplicates answers already in your body, because you end up with two near-identical chunks competing to represent the same idea. Retrieval systems dedupe and split ranking signal across near-duplicates, so you can hand the model the weaker copy and bury the stronger one you wrote in the article proper.
- Each answer is unique content that appears nowhere else on the page
- Every question is one real people actually ask, not filler from a generator
- Answers are self-contained, so a retriever can lift one cleanly
- It duplicates answers already in the body, creating near-identical chunks
- It's padded with keyword-shaped questions no real user searches
- It sits below the fold, adding nothing the article didn't already say
The other common damage is invented questions. FAQ generators pad a page with questions phrased for keywords rather than for humans, and those chunks never match a real prompt, so they add page weight and dilution without a single citation to show for it. A question no one asks is a chunk no one retrieves. The discipline that separates a quotable answer from empty filler — a specific fact instead of a vague gesture — is the sentence-level shift we walk through in GEO vs SEO writing.
How do you build FAQs that AI will use in 2026?
Start from the questions people actually ask you, answer each one once in a self-contained passage, and put those passages where they belong in the article rather than in an appendix. The single best source for "what do people actually ask" isn't a generator or a competitor's page — it's your own Google Search Console data, where the real question queries that already bring you impressions show exactly which how, does, can and why questions have demand.
That's the point where the work meets the product: Contexta's AI content editor is driven by your Search Console queries, so it rewrites and generates answer-first passages aimed at the real questions your pages already rank for, in your site's own voice, instead of inventing FAQ filler no one searches. It turns the FAQ question from "what should I pad the page with" into "which real question is this page not yet answering cleanly" — and that is the version of FAQ content assistants still reach for in 2026.
FAQ
Should I remove FAQ schema from my site in 2026?
No — there's no need to remove it, but there's no reason to invest in it either. FAQPage is still a valid Schema.org type and Google has said unused structured data doesn't cause problems, so leaving it in place is harmless. It just no longer earns a rich result after the May 7, 2026 deprecation and there's no confirmed evidence it wins AI citations, so treat it as low-priority and put your effort into the answer content instead.
Does FAQPage schema help me get cited by ChatGPT or Perplexity?
There's no confirmed evidence that it does — assistants pull clean question-and-answer content whether or not the markup is present. Google's own guidance states no special schema is required for AI Overviews or AI Mode, and the passage that gets lifted is chosen on the quality and self-containment of the text, not the tags around it. Write the answer to stand on its own and the markup becomes irrelevant to whether you're cited.
Where should FAQ content live — a block at the bottom or in the body?
In the body, as answer-first sections, because AI assistants chunk the whole page rather than privileging a region labelled 'FAQ'. A question-shaped heading with the answer in its first sentence produces the same liftable passage a bottom-of-page block would, and usually a stronger one. A dedicated FAQ block only helps when it answers something the body never covered, in unique, self-contained text.
How do I choose which FAQ questions to answer?
Use the questions people actually ask you, which your Google Search Console data reveals directly. The question-shaped queries that already bring your pages impressions — the real how, does, can and why searches — show which answers have genuine demand. Answering those cleanly beats padding a page with keyword-shaped questions from an FAQ generator, which create chunks that never match a real prompt.