Schema markup had a tough week. Google ended FAQ wealthy outcomes. 4 days later, Ahrefs revealed a report, discovering that including JSON-LD didn’t produce a transparent quotation carry throughout Google AI Overviews, AI Mode, or ChatGPT.
These developments weaken two frequent pitches for schema markup: elevated SERP visibility and potential AI quotation beneficial properties. This text examines their implications and what the info signifies about schema’s future.
Google’s Seen Schema Rewards Have Been Narrowing For Years
Google has been pulling again seen Search rewards tied to particular structured knowledge sorts since 2023. Google restricted FAQ wealthy outcomes to authoritative authorities and well being websites, and HowTo wealthy outcomes had been restricted to desktop and later deprecated.
In 2025, Google introduced the retirement of a number of structured knowledge options, together with Course Information, Declare Evaluate, and Estimated Wage. Guide Actions was initially included however later carved out after Google eliminated its deprecation banner. Google known as the remaining retirements “not generally utilized in Search” and not offering worth to customers.
In 2026, Apply Drawback structured knowledge was deprecated. John Mueller famous on Reddit that “markup sorts come and go, however a treasured few you must maintain on to.”
The sample is that seen structured knowledge rewards have disappeared after turning into acquainted search engine optimisation techniques. The markup itself stays legitimate, however the wealthy end result doesn’t. Google doesn’t all the time describe these removals as responses to overuse, however the sample gives much less purpose to deal with any single markup kind as a sturdy technique.
These latest updates differ as a result of the proof for one proposed substitute worth additionally weakened. The “GEO” advisory area claims schema boosts AI citations, and Ahrefs knowledge examined a part of that.
What The Ahrefs Report Discovered
Ahrefs tracked 1,885 net pages that added JSON-LD schema. Every web page was matched in opposition to management pages that by no means added schema. Quotation adjustments had been measured throughout Google AI Overviews, AI Mode, and ChatGPT.
The outcomes had been flat. Google AI Mode confirmed +2.4%, ChatGPT confirmed +2.2%, and Google AI Overviews confirmed -4.6%.
The primary two had been too small to inform aside from random variation. The AI Overviews decline was statistically important, however Ahrefs mentioned it will probably’t confidently attribute that to schema.
Each web page within the dataset already had greater than 100 AI Overview citations earlier than any schema was added. These pages had been already being crawled and cited.
Ahrefs acknowledged that for pages not but seen to AI, schema may nonetheless assist with crawling, parsing, or indexing. However their knowledge can’t affirm that.
Gianluca Fiorelli, a strategic search engine optimisation advisor, known as the examine “one of many extra trustworthy items of analysis to come back out of the AI Search area in 2026.” However he argued the scope was narrower than the headline steered. He in contrast it to “testing whether or not including a label to a bottle already on the grocery store shelf makes prospects choose it up extra usually.”
Ahrefs additionally cited a searchVIU experiment that discovered 5 AI programs relied on seen HTML throughout direct web page retrieval and didn’t use hidden JSON-LD, Microdata, or RDFa. That discovering covers one stage of the pipeline. It doesn’t rule out schema enjoying a task earlier in indexing or entity understanding.
Ryan Regulation, Ahrefs’ director of content material advertising, summarized the discovering on LinkedIn, saying:
“Does including schema markup assist your pages get cited in AI search? Most likely not,” he wrote. He added that schema is “most likely not some magic repair for bettering your AI citations.”
The Practitioner Debate
Each updates land in the course of an lively argument about schema and GEO.
Roughly 168,000 pages use the phrase “FAQ schema is important for GEO,” in accordance with search outcomes that Lily Ray, VP of search engine optimisation and AI Search at Amsive, flagged on LinkedIn. She known as the development acquainted.
“Something that may be spammed in search engine optimisation, might be spammed,” Ray wrote. She’d warned about this in a 2019 Moz article when FAQ schema first launched, and described Google’s FAQ elimination as the identical cycle repeating.
Ray hedged all through her put up, calling it “placing on my tin foil hat” and “simply an thought.” However the sample she described is identical one seen within the timeline above. A helpful markup kind will get scaled as a tactic, Google pulls the reward, and the business strikes on to the following one.
Joost de Valk, founding father of Yoast, made the connection specific in a weblog put up. “The GEO business is replaying early search engine optimisation, simply sooner,” de Valk mentioned. “And the FAQ schema deprecation is the primary concrete proof level that the cycle is again on.”
He additionally filed a Schema.org proposal for a brand new FAQSection kind to deal with what he sees because the structural downside, separating “this web page has an FAQ part” from “this web page IS an FAQ.”
The frustration was sharpest from practitioners who’d been watching the GEO playbook harden round schema as its most concrete advice. Mark Williams-Cook dinner, director at Candour and founding father of AlsoAsked, shared the Ahrefs report on LinkedIn.
“GEO bros are promoting snake oil with schema to spice up citations, however folks like Gianluca Fiorelli are speaking sense,” he posted.
Marie Haynes, founding father of Marie Haynes Consulting, commented on Ray’s put up with a distinct concept altogether.
“My concept is that Google wanted our FAQs to coach AI in order that they gave us incentive so as to add them (aka wealthy outcomes.) And now they don’t want them anymore,” she wrote. The speculation is unconfirmed by any main supply, however it reveals how far the hypothesis has traveled.
Some practitioners pushed again on the gloomier readings. Google’s broader steering nonetheless presents structured knowledge as a technique to make web page data machine-readable, and at a 2025 Search Central Dwell occasion in Madrid, the Search Relations crew informed practitioners that supported structured knowledge sorts are nonetheless value utilizing.
What The Knowledge Can’t Reply But
Whether or not schema helps pages that aren’t but being cited is a separate query that the info can’t reply, as a result of each web page already had greater than 100 AI Overview citations earlier than schema was added.
The check additionally pooled all schema sorts collectively. Article, FAQ, Product, HowTo, and Group had been all handled as one class. Sort-specific results haven’t been remoted, and so they may look totally different.
The 30-day measurement window could miss slower results, and on reside web sites, schema adjustments can overlap with different web page adjustments, making it onerous to separate what schema did from what modified round it. The report solely examined schema within the web page’s HTML, not schema injected by way of JavaScript, which AI crawlers deal with otherwise.
Ahrefs measured Google AI Overviews, AI Mode, and ChatGPT. Whether or not Bing, Copilot, Perplexity, Claude, or different reply programs deal with schema otherwise from the programs Ahrefs measured is an open query.
Google’s FAQ deprecation discover says the corporate will proceed utilizing FAQ structured knowledge to “higher perceive” pages. What that produces in measurable phrases is unclear. The identical uncertainty applies as to whether schema impacts citations not directly, by way of eligibility, entity understanding, or supply choice, moderately than throughout the direct retrieval that searchVIU examined.
No one has revealed knowledge that isolates that path.
Why This Issues
The Ahrefs knowledge provides no measured purpose so as to add JSON-LD, anticipating short-term AI quotation beneficial properties for pages already seen in AI Overviews. The trickier query is what to do with schema methods extra broadly.
Product, Evaluate, Occasion, Video, and another structured knowledge sorts nonetheless help lively wealthy end result options. Group, Particular person, and Article markup can nonetheless assist describe entities and content material, even when the payoff is much less seen.
A blanket “schema doesn’t work” studying overstates what the info confirmed, as a result of the check pooled every kind and measured just one final result. What the info does problem is a selected gross sales pitch.
“Add schema to spice up AI citations” has been one of many extra concrete suggestions in GEO guides. For instance, Frase.io known as schema markup “critically essential for AI search, GEO, and AEO.”
With out knowledge help for that declare, it’s more durable to justify the funding. AI programs in searchVIU’s check relied on seen HTML throughout retrieval, not JSON-LD. That implies content material construction, clear headings, and direct solutions in prose could matter extra for AI quotation than markup construction.
Wanting Forward
The query hanging over the search engine optimisation business is the place schema creates measurable worth. Including JSON-LD didn’t measurably enhance AI citations for pages already seen in AI Overviews.
For these pages, schema seems to be extra like plumbing that serves different programs than a lever that strikes quotation counts. That’s nonetheless actual worth, however it’s a distinct pitch.
Featured Picture: BEST-BACKGROUNDS/Shutterstock
