Ahrefs added schema markup to 1,885 pages and tracked them against roughly 4,000 control pages. AI Overview citations fell 4.6% relative to the control pages. AI Mode and ChatGPT moved within noise.
That is the largest controlled test I found on the question, and it points the wrong way for anyone selling schema as an AI visibility service.
Schema still belongs on your site. It does one job, it does it well, and that job is not earning citations.
What schema markup actually is
A block of code that labels what a page is about, in a format machines read.
It usually sits in the page source as JSON-LD, inside a script tag. A human sees your homepage and understands that Acme sells scheduling software. A machine sees text. Schema tells it directly: this is an Organization, its name is Acme, its website is acme.com, and these are its profiles elsewhere.
The vocabulary comes from schema.org, which search engines helped build. There are hundreds of types. You need two or three.
Does schema markup help AI citations?
No. The best evidence I found shows no lift.
The Ahrefs test above is the controlled one. Google’s own guidance says you don’t need it. In Google’s words: “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.” I quoted that guidance in full when I wrote about llms.txt versus robots.txt.
Teams keep shipping it anyway. In my crawl of 3,877 US B2B software companies with 100+ employees, 247 of 3,236 homepages carry FAQPage markup in the HTML the server sends, and 423 carry SoftwareApplication or Product markup. Somebody billed for a share of that.
Teams keep shipping schema types that do not earn citations
Homepages carrying each markup type, out of 3,236 B2B software homepages. Each bar is a separate count against the full base, and the two counts can overlap.
Neither type is a route to AI citations. If you already have FAQPage markup, leave it. Do not add more.
Source: Scott King crawl of B2B software homepages, read from the server HTML before JavaScript runs, Oct 29, 2026 (two verified crawl passes). Base: 3,236 homepages.
The one job schema does well
It tells a machine which company you are. That is entity disambiguation, and it is worth an afternoon.
Three companies share your name. One is a staffing agency in Ohio. One is a defunct hardware brand with a Wikipedia page. You are the third. An answer engine assembling a response has to decide which of you the question refers to, and nothing on your page makes that obvious.
Organization markup with accurate sameAs links resolves it. The markup states your canonical name, and sameAs points to the profiles that corroborate it: LinkedIn, Crunchbase, your G2 listing, Wikidata if you have an item.
In my crawl, 2,140 of 3,236 homepages carry Organization schema in the HTML the server sends. 1,096 do not. If you are in that second group, this is the one piece of markup worth adding, and it takes a developer twenty minutes.
A third of B2B homepages carry no Organization schema
3,236 B2B software homepages, split by whether the server HTML carries Organization schema. Each block’s area is to scale.
If you are in the 1,096, Organization markup with accurate sameAs links is the one piece of schema worth adding.
Source: Scott King crawl of B2B software homepages, read from the server HTML before JavaScript runs, Oct 29, 2026 (two verified crawl passes). Base: 3,236 homepages.
What is an entity, in practice?
A thing a machine can identify and connect facts to, rather than a string of characters it happens to have read.
“Acme” is a string. The company Acme, founded in 2019, headquartered in Austin, selling scheduling software, with a LinkedIn page and 40 employees, is an entity. The difference matters because retrieval systems attribute claims to entities. I have not seen a system credit a claim it cannot tie to a name.
This is also why a vague homepage headline costs you twice. The sentence does not differentiate you, and it does not name you either.
How do you check your own entity?
Open six tabs and compare what each one says.
- Your homepage source, for the
Organizationblock and itsnamefield. - Your footer copyright line.
- Your LinkedIn company page.
- Your Crunchbase profile.
- Your G2 or Capterra listing.
- Wikidata, if an item exists.
Then check four facts across all six: company name, founding year, headcount range, and what you say you do. Contradictions are the finding. A machine has no way to resolve which source is right, so it may pick the one you like least.
Watch for one specific failure. Several engines state the same fact about you, a founding year say, and your own site never states it anywhere. That fact is probably propagating from a single indexed source. Correct it there, not on your website.
Two more things to confirm while you are in the source. Your schema must be server-rendered, because most AI crawlers never run your JavaScript. And every sameAs URL must resolve. A broken sameAs points a machine at nothing.
The author half nobody does
Your articles have a byline. Your markup probably does not.
If you publish under a named author, that author is an entity too. Three things do the same disambiguation job for a person that Organization does for a company. Person markup with sameAs pointing to a real LinkedIn profile. A visible bio with a stated role. A resolvable author page.
The common defect is a corpus published under “Team” or a house byline. It removes the only entity a retrieval system could connect the claims to. It is usually a template setting rather than a decision anyone made.
One caution, because this gets oversold. Author markup is disambiguation, not an expertise score. I have not seen a test showing it lifts citations. It makes you identifiable, which is the floor, not the lever.
What to skip
Three things, and each one gets sold as AI visibility work.
FAQPage and HowTo schema for citations. Rich results for both are gone and the Ahrefs test, which pooled every schema type including FAQ and HowTo, found no uplift. If you already have FAQPage markup, leave it. Do not add more.
Schema volume as a score. Nobody is counting your types. A site with twelve schema types and a generic headline loses to a site with two types and a defensible claim.
Markup as a substitute for text. A claim that exists only in structured data and not in readable page text is a claim a human reader never sees. Write the sentence, then describe it in markup if you like.
The rest of the refused list sits here, with the study behind each one.
What actually correlates
Mentions, not markup.
Ahrefs measured branded web mentions at 0.664 correlation with AI Overview presence, against 0.218 for backlinks. Correlation is not causation, and Ahrefs says so. But the ranking of factors is the useful part: being talked about elsewhere tracks with being named, and no markup substitutes for it.
So spend the markup budget once, correctly, then move on to the work that actually changes the count.
Ask your developer this
Two questions, and both are answerable in a minute. Is Organization markup in the server-rendered HTML, and does every sameAs URL in it resolve?
If the answer is no and yes-ish, you have twenty minutes of work and nothing more to do with schema this quarter.
FAQ
Does schema markup help AI search?
No. The best evidence I found shows no lift. Ahrefs added schema markup to 1,885 pages and tracked them against roughly 4,000 control pages. AI Overview citations fell 4.6% relative to the control pages. AI Mode and ChatGPT moved within noise.
Do AI crawlers read schema added with JavaScript?
Your schema must be server-rendered, because most AI crawlers never run your JavaScript.
How common is Organization schema on B2B software sites?
In my crawl, 2,140 of 3,236 homepages carry Organization schema in the HTML the server sends. 1,096 do not.
Does FAQ schema still work?
If you already have FAQPage markup, leave it. Do not add more. In my crawl of 3,877 US B2B software companies with 100+ employees, 247 of 3,236 homepages carry FAQPage markup in the HTML the server sends. Rich results for FAQPage and HowTo schema are gone.
What matters more than schema for AI Overviews?
Mentions, not markup. Ahrefs measured branded web mentions at 0.664 correlation with AI Overview presence, against 0.218 for backlinks. Correlation is not causation, and Ahrefs says so.
The audit checks entity markup and entity resolution, including whether your name matches across your site, LinkedIn and Crunchbase. It pairs that with 150 observations across ChatGPT, Google AI Mode and Perplexity, so you can see whether the engines know who you are today.
Scott King helps B2B technology companies build AI marketing systems and create differentiated content.