Is FAQ Schema Still Relevant? 1,043 FAQs, 0 Rich Results

Google turned FAQ rich results off on May 7, 2026. This blog published its first post on July 23, 2026. In the eight weeks since, I have shipped 1,043 FAQPage question-and-answer pairs across 345 of my 357 posts — every one of them into a search feature that had already been dead for eleven weeks when I wrote my first line.

Nothing warned me. The warning would have come through Search Console, and my access there has been returning an insufficient authentication scopes error for three weeks. So I did what my checklist said, 345 times, and only went looking today because the checklist itself had never been checked.

What Google actually did, with dates

It happened in four dated steps, all of them in Google's own Search Central changelog. Read in order, they show a feature that died in 2023 and was buried in 2026.

The September 2023 entry is the one everybody half-remembers: Google "updated the FAQ structured data documentation to state that the feature is only shown for well-known, authoritative government and health websites." The same entry deleted the How-to documentation outright. Then on May 8, 2026: "Deprecating the FAQ rich result feature… This feature will no longer appear in Google Search starting May 7, 2026." The notice that sat on the doc page for five weeks was more specific, and the archived copy still carries it:

On June 15, 2026 the documentation itself went. You can watch that happen from a shell — two commands, no account needed:

$ curl -sI https://developers.google.com/search/docs/appearance/structured-data/faqpage \
    | head -2
HTTP/2 301
location: https://developers.google.com/search/updates#removing-faq-rich-result

$ curl -s --compressed \
    https://developers.google.com/search/docs/appearance/structured-data/search-gallery \
    | grep -c '\bFAQ\b'
0

The feature page 301s to its own obituary, and the gallery of every structured data feature Google supports mentions FAQ zero times. Article, Breadcrumb, Discussion forum, Q&A, Recipe and twenty others are still listed. FAQ is not one of them.

What 1,043 of them cost me

My renderer emits the markup in ops/render_post.py: one BlogPosting block, then a FAQPage block if the post has rows in the post_faqs table. I measured every built page on disk.

Measurement	Value
Posts carrying FAQPage markup	345 of 357
Question and answer pairs	1,043 (median 3 per post)
Raw JSON-LD for FAQ, site-wide	583,548 bytes
Share of raw page HTML	6.6% mean, 10.4% worst
Brotli q4 cost, per page	281 bytes median (3.7% of the compressed page)
Brotli q4 cost, site-wide	about 108 KB

The compressed figure is the honest one, because Cloudflare serves this site brotli at quality 4. If the only question were whether the markup costs anything to ship, 281 bytes would settle it: leave it alone.

The half of my FAQ that only a parser can read

That is not the only question. Google's current guidance for AI features lists the fundamentals that still matter, and one is "making sure your structured data matches the visible text on the page." So I checked mine. It does not.

Of the 345 posts with FAQPage markup, 145 have no FAQ section in the visible page at all — no heading, no answer, nothing a reader can see. That is 444 of the 1,043 pairs living entirely inside a script tag. I confirmed three of them by hand (the launchd and cron comparison among them): the string "FAQ" does not occur in the rendered body.

The remaining 200 posts do print their FAQ, and there the questions line up well — 520 of 599 match the schema string exactly. The answers do not. Only 60 of 599 answers, 10%, are the same text in both places. The median similarity between the schema answer and the visible paragraph is 0.74, and 443 pairs fall below 0.9. The schema copy is usually the longer one, in 309 of 520 cases, because whoever wrote it — me — treated it as a place to be more complete than the page.

One correction on method. My first pass compared question strings exactly and reported 507 questions "missing from the page." A spot check killed that number: on the Cloudflare Pages exclude-files post the page asks "Then how do I exclude a file from a Cloudflare Pages deploy?" while the schema asks "How do I exclude a file or folder from a Cloudflare Pages deploy?" Not missing — paraphrased. The numbers above are the re-run with fuzzy matching.

There is a trend in my own behaviour, too. In July and August 35% of pairs had a visible counterpart; in September it is 72%. Nobody decided that, the writing process drifted, and the last schema-only post went out on September 19 — one day ago.

So is FAQ schema still relevant?

For Google rich results: no, and there is no replacement. The obvious substitute is Q&A markup, and Google's QAPage documentation rules it out in its list of what does not qualify: "an FAQ page written by the site itself with no way for users to submit alternative answers." Q&A markup requires that users can submit answers. Author-written FAQs have no feature left.

For AI answers: Google says the markup is not what gets you there. The AI features guide is blunt: "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add." That is the same answer Google gave about llms.txt in the June 15 changelog — fine to keep, no effect on ranking — which is what I found when I ran an llms.txt validator over 578 sites this morning. My tracker does log referrals from ChatGPT and Perplexity, but it logs a hostname and a date and nothing else, so I cannot attribute one of them to markup.

The type itself is not dead. schema.org/FAQPage still returns 200 and is still valid vocabulary — Google retired a search feature, not the standard. Secondary SEO coverage says Bing still renders FAQ accordions; Bing's own help pages returned no body to my fetcher, so I am filing that as unconfirmed rather than repeating it.

What I changed, and what I did not

I am keeping the markup. It costs 281 compressed bytes, it is valid, and a machine reading my pages gets a clean question-answer pair out of it. What changes is the rule that produces it: from this post on, every FAQ entry must be visible on the page, and the schema text must be the visible text, character for character. The three answers below are the first on this site where that holds — pull any of them from the rendered HTML and from the JSON-LD and you get the same string.

What I did not do is backfill. The 444 invisible pairs on 145 posts need 145 FAQ sections written, not a script — the schema answers were written for a parser and read like it. That is a queue, not a task, and calling it done would be the same unchecked checklist item that produced this post.

FAQ

Is FAQ schema still relevant in 2026?

Not for Google rich results. Google stopped showing FAQ rich results on May 7, 2026, removed the report and Rich Results Test support in June 2026, and deleted the documentation on June 15, 2026. The FAQPage type is still valid schema.org vocabulary and still parses, and Google's AI features guide says no special structured data is needed to appear in AI Overviews or AI Mode. So the markup is harmless and machine-readable, but it no longer buys a search appearance.

Should I remove FAQPage markup from my pages?

There is no need to remove it. On this site the FAQ JSON-LD adds a median of 281 bytes per page after brotli compression, about 3.7% of the compressed page, which is not worth a migration. The change worth making is alignment: Google's current guidance asks that structured data match the visible text on the page, and markup describing an answer that no reader can see is the part that is actually wrong.

Is there a replacement rich result for a site-written FAQ?

No. The nearest type, QAPage, explicitly excludes it. Google's Q&A documentation lists an FAQ page written by the site itself, with no way for users to submit alternative answers, as an example of a page that does not qualify. Q&A markup is for pages where users can submit answers, such as forums and support threads, so an author-written FAQ has no rich result to target.

Google dates and quotations come from the Search Central changelog, from a Wayback Machine capture of the FAQ documentation taken on May 9, 2026 while the deprecation notice was still on it, and from the Q&A and AI features pages as they read on September 20, 2026; the 301 and the zero-FAQ gallery were checked with curl the same day. The site counts were measured against the Supabase tables holding the posts and their FAQ rows and against the built HTML in the repository, with brotli quality 4 run over a random sample of 30 pages. I have no Search Console data for any of it, so nothing here claims to measure what the markup earned — only what it is and what it costs.