Schema markup has had a rough week. Google FAQ completed, rich results. Four days later, Ahrefs published a reportnoting that by adding JSON-LD did not produce a clear increase in citations on Google AI InsightsAI or ChatGPT mode.
These developments weaken two common arguments for schema markup: increased SERP visibility and potential AI citation gains. This article examines their implications and what the data indicates about the future of the scheme.
Google’s visible schema rewards have been shrinking for years
Google removed visible search rewards linked to structured data types since 2023. Google FAQ restricted to rich results on authoritative government and health sitesand rich HowTo results were limited to desktop and later deprecated.
In 2025, Google announced the retirement of several structured data featuresincluding course information, claims review and estimated salary. Book Actions was initially included, but was later removed after Google removed its deprecation banner. Google called the remaining pensions “not commonly used in research” and no longer provides value to users.
In 2026, structured data Practice Problem was obsolete. John Mueller noted on Reddit that “markup types come and go, but you should keep a few valuable ones.”
The trend is that the visible rewards of structured data have disappeared after becoming familiar SEO tactics. The markup itself remains valid, but the enriched result does not. Google doesn’t always describe these removals as responses to excessive usage, but this model offers less reason to consider a single type of markup as a sustainable strategy.
These recent updates differ because the evidence for a proposed replacement value has also weakened. The “GEO” advisory space complaints pattern drives AI citations, and Ahrefs data tested some of that.
What the Ahrefs report found
Ahrefs tracked 1,885 web pages adding JSON-LD schema. Each page was compared to control pages that never added schema. Citation changes were measured in Google AI Previews, AI Mode, and ChatGPT.
The results were flat. Google AI mode displayed +2.4%ChatGPT showed +2.2%and Google’s AI previews showed -4.6%.
The first two were too small to be distinguished from random variation. The drop in AI insights was statistically significant, but Ahrefs said he couldn’t definitively attribute it to the pattern.
Each page of the dataset already contained over 100 AI Overview citations before a schema was added. These pages were already explored and cited.
Ahrefs recognized that for pages not yet visible to the AI, schema can still help with crawling, analysis, or indexing. But their data cannot confirm this.
Gianluca Fiorelli, strategic SEO consultant, called the study “is some of the most honest research to come out of the AI research space in 2026.” But he argued that the scope was narrower than the title suggested. He compared it to “testing whether adding a label to a bottle already on the shelves of a supermarket encourages customers to pick it up more often”.
Ahrefs also cited a researchVIU experience which found that five AI systems relied on visible HTML during direct page retrieval and did not use hidden JSON-LD, microdata, or RDFa. This finding covers one stage of the pipeline. This does not exclude that the schema plays a role earlier in indexing or understanding of the entity.
Ryan Law, Director of Content Marketing at Ahrefs, summarized the conclusion on LinkedIn, saying:
“Does adding schema markup help your pages get cited in AI search? Probably not,” he wrote. He added that the scheme is “probably not a magic bullet to improve your AI citations.”
The practitioners’ debate
Both updates land in the middle of an active discussion about the scheme and GEO.
Approximately 168,000 pages use the phrase “FAQ Schema is essential for GEO,” according to research findings that Lily Ray, VP of SEO and AI Search at Amsive, reported on LinkedIn. She called the trend familiar.
“Anything that can be spammed in SEO will be spammed,” Ray wrote. She had warned him in a Moz Article 2019 when the FAQ scheme was first launched and described Google’s removal of the FAQ as the same cycle repeating itself.
Ray hedged throughout her post, calling it “putting on my tin foil hat” and “just an idea.” But the pattern she described is the same one seen in the timeline above. One type of useful markup is scaled as a tactic, Google gets the reward, and the industry moves on to the next one.
Joost de Valk, founder of Yoast, made the link explicit in a blog post. “The GEO industry is replaying the early days of SEO, but faster,” de Valk said. “And the deprecation of the FAQ schema is the first concrete evidence that the cycle is back.”
He also submitted a Schema.org proposal for a new FAQSection type to fix what it sees as the structural problem, separating “this page has an FAQ section” from “this page IS an FAQ”.
Frustration was greatest among practitioners who had seen the GEO manual harden around the diagram as the most concrete recommendation. Mark Williams-Cook, director of Candor and founder of AlsoAsked, shared the Ahrefs report is LinkedIn.
“The GEO brothers are selling snake oil with a scheme to increase citations, but people like Gianluca Fiorelli are talking common sense,” he posted.
Marie Haynes, founder of Marie Haynes Consulting, commented on Ray’s post with a completely different theory.
“My theory is that Google needed our FAQs to train the AI, so they tricked us into adding them (i.e. rich results.) And now they don’t need them anymore,” she wrote. The theory is not confirmed by any primary sources, but it shows how far the speculation has traveled.
Some practitioners rejected the darker readings. Google’s broader guidelines still present structured data as a way to create information on pages. machine readableand to a Search Central Live 2025 event in MadridThe Search Relations team told practitioners that the supported structured data types are still worth using.
What the data can’t yet answer
Whether the schema helps pages that are not yet cited is a separate question that the data cannot answer, because each page already contained over 100 AI Overview citations before the schema was added.
The test also grouped all types of diagrams. Article, FAQ, Product, How-To, and Organization were all treated as one category. The specific effects of each type have not been isolated and could be different.
The 30-day measurement window can miss slower effects, and on live websites, schema edits can overlap with other page edits, making it difficult to distinguish between what the schema did and what changed around it. The report only looked at the schema contained in the page’s HTML, not the schema injected via JavaScript, which AI crawlers process differently.
Ahrefs measured Google AI Previews, AI Mode, and ChatGPT. Whether Bing, Copilot, Perplexity, Claude, or other response systems treat the pattern differently than the systems measured by Ahrefs is an open question.
Google’s FAQ deprecation notice states that the company will continue to use structured data from the FAQ to “better understand” the pages. What this produces in measurable terms is unclear. The same uncertainty applies as to whether schema affects citations indirectly, via eligibility, entity understanding, or source selection, rather than in the direct retrieval tested by searchVIU.
No one has published data that isolates this pathway.
Why it matters
The Ahrefs data gives no measured reason to add JSON-LD, expecting near-term AI citation gains for pages already visible in AI previews. The trickier question is what to do with schema strategies more generally.
Product, Reviews, Event, Video, and some other structured data types still support active rich results features. Organization, person, and article markup can still help describe entities and content, even when the payoff is less visible.
A blanket reading of “the pattern doesn’t work” overestimates what the data shows, because the test lumped all types together and only measured one outcome. What the data calls into question is a specific sales pitch.
“Add schema to increase AI citations” was one of the GEO Guides’ most concrete recommendations. For example, Frase.io called schema markup “critically important to AI search, GEO, and AEO.”
Without data to support this claim, it is more difficult to justify the investment. The AI systems in the searchVIU test relied on visible HTML during retrieval, not JSON-LD. This suggests content structureclear titles and direct prose responses may matter more to AI citation than markup structure.
Looking to the future
The question hanging over the SEO industry is where schema creates measurable value. Adding JSON-LD did not measurably increase AI citations for pages already visible in AI previews.
For these pages, the diagram looks more like plumbing that serves other systems than a lever that moves citation counts. It’s still a real value, but it’s a different argument.
Featured Image: BEST STORIES/Shutterstock
