Weekly field note

What Schema Markup Still Buys After Google Retired the FAQ Rich Result

Published November 9, 2026

Google requires no markup for its AI features, a matched test of 1,885 pages measured no citation lift, and the remaining case rests on explicit facts and rich results

Structured data about a creative work

*Image: Structured data about a creative work by Sandra Fauconnier, licensed under CC BY-SA 4.0.*

AEO Mastery, the course that replaced GeoAnalyzer's audit scanner, keeps structured data on its priority list. The public summary of its schema lesson states that schema markup is "not required for AI search," that Google specifies no special schema for its AI features, and that FAQ and HowTo markup no longer produce Google rich results. A team planning a month of AEO work now has to decide what the remaining case is worth, and how to check it.

The course homepage lists 8 modules and 32 lessons and displayed a $25 price against a $59 list price when checked on November 9, 2026. The lesson and Google's documentation agree that AI features need no special markup. The evidence that schema raises AI citations is thinner than the correlation statistic that Ahrefs says gets shared in LinkedIn carousels and conference slides as proof that schema is an AI visibility lever.

What the documentation sets as the bar

Google's guide to AI features says a page needs to be indexed and eligible to appear in Search with a snippet, and that "there are no additional technical requirements." Its May 2026 guide to optimizing for generative AI features adds one condition on top: a site "must be included in Search generative AI features in Search Console to be eligible for display" in those features.

The May guide's mythbusting section lists "LLMS.txt files and other 'special' markup" among the things site owners can ignore, because Google Search itself doesn't use them. The AI features page puts the schema point plainly: "There's also no special schema.org structured data that you need to add." searchVIU reports that the May guide, as of July 10, 2026, describes structured data as not required for generative AI search and still recommends it for rich results. The AI features documentation also asks that structured data match the visible text on the page.

Google's general structured data guidelines set stricter conditions on eligibility and accuracy. A structured data issue can trigger a manual action that costs a page its rich result eligibility, violating a quality guideline can keep syntactically correct markup from being displayed, appearance is never guaranteed, and the data has to be "a true representation of the page content."

What Google stopped showing

Google stopped showing the FAQ rich result on May 7, 2026. The Search Central changelog records a deprecation notice dated May 8 and a June entry that removed the FAQ rich result documentation because the feature "is no longer shown in Google Search results." The AEO Mastery lesson summary adds that HowTo markup no longer produces Google rich results either.

That changes what an older schema audit measures. Any FAQPage coverage it counted now describes markup Google does not display. A Q&A section on the page can still give a reader a fast answer to a question.

The matched test

The strongest published attempt to separate markup from the sites that install it came from Ahrefs in May 2026. It began with a scan of 6 million URLs and found that pages cited by AI were almost three times as likely to carry JSON-LD as pages that were not. Ahrefs cautions that the gap "could easily have been correlation, not cause," since schema tends to live on better-maintained, more technically sophisticated sites that also publish stronger content, build authority and earn links.

To isolate the markup itself, the study tracked 1,885 pages that added JSON-LD between August 2025 and March 2026 and compared them against 4,000 control pages from other domains that had similar citation levels and never added markup. A difference-in-differences comparison across Google AI Overviews, AI Mode and ChatGPT produced three results. AI Overview citations on treated pages fell 4.6 percent against controls, a gap the post calls small but statistically significant. In absolute terms that is an average loss of around 12 daily citations per page, in a sample where most pages were getting hundreds. Both groups were already declining before the change, and Ahrefs leaves the cause of the gap open. AI Mode gained 2.4 percent and ChatGPT 2.2 percent, both "statistically indistinguishable from zero." "We can't tell whether the schema did a tiny bit of good or nothing at all," the post says.

Ahrefs lists several limits on what the test shows. Pages that add JSON-LD often change other things at the same time, such as links, content and technical fixes, and the study cannot fully separate schema from those changes. Every page in the dataset had at least 100 AI Overview citations in February 2025, before any schema was added, so the comparison describes pages AI systems already retrieved. All schema types were pooled into one treatment. The measurement window was 30 days. Only JSON-LD in the page's HTML was tracked, leaving out schema injected via JavaScript.

The retrieval test

searchVIU ran a separate test on October 30, 2025, and published it that December. The test page carried eight variants of a fictitious product, each with its own price, and placed each price in exactly one source: visible HTML, JavaScript, JSON-LD only, JSON-LD inserted by JavaScript, hidden microdata, visible microdata, conflicting hidden markup, or visible RDFa. ChatGPT, Claude, Gemini, Perplexity and Google AI Mode were then asked for the prices.

No system named the price that existed only in JSON-LD, and none named the conflicting pair. Every price that came back came from visible HTML or from the price inserted by JavaScript. Gemini was the only system to read the JavaScript price during a direct fetch. Perplexity and Google AI Mode answered after the page entered their indexes. Perplexity named only the JavaScript-inserted price, and AI Mode named that one plus a visible price.

The post distinguishes four routes by which markup can reach a model: training data, an index, search retrieval, and a live fetch. Its test mainly covers the live fetch, and it states that the test does not answer whether Google's AI Overviews uses structured data from Google's index. It also notes that the experiment has not been repeated since October 2025, and its list of common mistakes includes drawing conclusions about all systems and all points in time from a single test.

What the tests imply for a page

Read together, the documentation and the two tests separate three jobs that a single schema score blends:

The third job is where the searchVIU result lands. In its single October 2025 test, a price that existed only in JSON-LD reached none of the five AI systems, and searchVIU lists keeping important facts only in JSON-LD and off the page as a common mistake. Googlebot does read JSON-LD for rich results, and the test did not check what AI Overviews take from the index, so the finding covers live fetches by those five systems on that date. A team that wants a fact to reach them should put it on the page as well as in the markup.

What to write, and how to check it

The AEO Mastery lesson summary recommends prioritizing Organization, WebSite, Article, Product and LocalBusiness, and advises linking entities with @id. Neither test compares schema types: the matched study pooled them, and the retrieval test found no system reading a price that lived only in product JSON-LD, which is an argument for showing the price on the page. Take the list as the course's view of which facts matter for each site type, and run your own test to learn whether the markup moved a citation on your pages.

Two practices follow. Keep the markup describing what is visible, and validate it, because invalid markup loses rich result eligibility before it can do anything else. Then test the citation question on your own pages instead of assuming either answer, with the design the Ahrefs study describes: pick five to ten pages already receiving some citations, match them against five to ten similar pages left untouched, record baseline counts across AI Overviews, AI Mode and ChatGPT, change only the markup, and compare the two groups after 30 days or more.

Whether Google's AI Overviews reads structured data from its own index remains untested by either study. Ahrefs notes that for pages AI systems are not yet seeing, schema "might still play a role" in getting them crawled, parsed or indexed, and that its study cannot speak to that, so the answer could differ for pages outside the retrieval set.

Write schema when a fact needs to be stated explicitly for a machine and shown to a reader, and when a rich result depends on it. Do not budget it as a citation lever for pages that AI systems already retrieve, because the matched test found no clear lift there.