Does Schema Markup Help You Get Cited by AI?

Every SEO vendor selling AI-visibility services right now leads with some version of the same pitch: get your schema markup right, and you'll get cited by ChatGPT and Google's AI Overviews. It's in most generative engine optimization sales decks I've seen this year. GEO is what the category calls itself. I think it's wrong, and I'd rather show you why I believe that than just assert it.

Schema markup is structured data you put in a page's code. It tells a machine "this is a product, this is the price, this is the review count" without making it guess from the prose. It's real, it's useful, and it's been part of good technical SEO for over a decade. None of that is in question. What's in question is the specific claim being sold right now, that adding or upgrading schema markup gets you cited by AI answer engines. That claim has been tested, twice, and it failed both times.

The first refutation came from inside a vendor's own stated findings, the kind of source that has every incentive to make schema look like the lever and still didn't find one. The second came from a controlled test built specifically to isolate the question: does changing the structured data around a page change whether it gets cited, holding the actual content constant. It didn't move the needle. Formatting-only edits, and schema markup falls squarely into that category, show no measured impact on AI citation.

While we're on the subject of things the industry sells that don't hold up: "authoritative tone" doesn't do anything measurable either. Writing your page in a more confident, more official-sounding register, the kind of copy an agency charges a premium for, doesn't correlate with getting cited. The machine isn't grading your delivery. It's grading whether your passage actually answers the question in front of it.

So what does correlate with citation? Several things, and none of them are a code snippet you paste in once and forget about.

Page quality, in the plain sense a person would recognize walking through the content. Clarity: can a reader, human or machine, tell exactly what you're claiming and what it's based on. Fluency and readability: is the passage well-written prose, not fragmented keyword salad stitched together for a crawler. And semantic alignment to the specific question: does your passage match the actual sub-question a customer is asking, not just the broad topic around it. I want to be straight with you about the strength of these findings. The direction, that quality and clarity beat decoration, is well supported across the research I trust. The exact size of the effect is contested, and I'm not going to hand you a number I can't stand behind. Anyone quoting you a precise percentage lift for this kind of thing is doing the same move the schema vendors are doing: selling certainty a genuinely uncertain field hasn't earned yet.

It's worth asking why schema became the pitch in the first place, because the answer says something about the whole GEO category. Schema is easy to sell. It's a defined technical task with a start and an end, an agency can quote you a price for it, install it, and hand you an invoice. "Rewrite your content to actually answer the specific question a customer would ask" doesn't fit that model nearly as well. It's slower, it requires someone who understands your business and not just your codebase, and it's harder to prove you did in a single line item. None of that makes schema the wrong technical practice. It just means the sales incentive and the actual evidence pulled in different directions, and the sales incentive won the pitch decks.

There's one more factor worth naming on its own: evidence density. Pages that carry real numbers, direct quotes, actual data a machine can lift into a generated answer, appear to do better than pages that describe things in general terms. Again, the direction here is reasonably well supported. The magnitude is not settled, and I'd rather tell you that plainly than round it up into a stat that sounds better in a slide.

Now, to be fair to schema markup, because I don't want this to read as "throw it out." Keep doing it. It's still worth the engineering time, just not for the reason vendors are charging you for. Schema still earns you rich results in classic Google search, the star ratings and price ranges that show up directly in a search listing. It still helps disambiguate your entity, telling a system unambiguously "this business, this location, this category" instead of leaving it to infer from context. And it's still good technical hygiene for classic organic search generally. Those are real, defensible reasons to keep clean schema on your site. "It gets you cited by AI" just isn't one of them, at least not based on anything I've seen actually tested.

Here's the real point of this post, and it's bigger than schema. The GEO industry, the whole generative engine optimization category that's sprung up around AI search, is selling certainty it hasn't earned. Schema-as-citation-lever is one specific claim I can point to and say: tested, twice, didn't hold. Watch for the pattern behind it. A vendor packages a technical checkbox as the secret to AI visibility, because a checkbox is easy to sell and "write a clear answer to the real question" isn't.

I'd rather give you the honest, less exciting version. There's no code you paste in that gets you cited. What correlates is the unglamorous stuff: write a real, clear, well-matched answer to the actual question your customer is asking, and back it with specifics instead of decoration. That's slower to sell as a service and harder to package into a dashboard, which is exactly why almost nobody's selling it to you straight.

Grade the evidence yourself before you buy the next "AI visibility" package. Ask three questions before you sign anything. What was actually tested, not just claimed. Was the test controlled for content quality, or did the vendor just point at a site that happens to have both good schema and good content and credit the wrong one. And would the finding survive someone trying to break it, or does it collapse the moment you ask for the underlying data. If the answer to any of those is "trust us," that's your answer.

Schema won't get you cited. A real answer to a real question will.