Schema markup is structured data, usually written in JSON-LD, that labels the content of a page for machines using the schema.org vocabulary. It is often sold as a way to earn citations in ChatGPT, Perplexity, and Google AI Overviews. The documented evidence supports a narrower claim, and a buyer should know where it stops.
- Google’s AI features documentation states there is no special schema.org structured data to add for AI Overviews or AI Mode.
- A Search Atlas analysis published December 14, 2025 concluded schema has no measurable effect on LLM visibility across OpenAI, Gemini, and Perplexity.
- That analysis did not publish its sample counts, and Search Atlas sells AI visibility software, so treat it as indicative, not definitive.
- Schema still has a job: Google’s structured data guidance ties it to search features, and Google asks that markup match visible page text.
- Do not pay for schema as a citation guarantee.
What vendors claim
The usual pitch has three steps. Add schema, so machines understand your pages, so language models cite you more. The first link is true by definition: markup is machine-readable. The second and third are separate empirical claims, and each needs its own proof.
That distinction matters for procurement. Schema can be valuable for classic search features while doing nothing for AI citations. A proposal that blurs the two is selling a result it has not shown. The reader implication is simple: ask which outcome the markup is meant to change, and where the measurement for that outcome comes from.
What Google says
Google’s documentation on AI features in Search is direct. It states that there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary. It adds that there is no special schema.org structured data you need to add, and that you don’t need new machine-readable files, AI text files, or markup to appear.
The only structured-data advice in that page is general: make sure your structured data matches the visible text on the page. For a buyer, that is a clear baseline. If a vendor says Google AI features require a particular schema type, the claim contradicts Google’s own published guidance and should be backed by a source. The same logic applies to llms.txt, which Google’s wording also covers.
What the Search Atlas analysis found
Search Atlas, an SEO software company, published an analysis on December 14, 2025 titled “The Limits of Schema Markup for AI Search.” It checked each page’s HTML for JSON-LD, microdata, or RDFa, computed each domain’s share of URLs with schema, and sorted domains into five buckets from no schema to full schema. It then compared LLM visibility scores across the buckets for OpenAI, Gemini, and Perplexity.
Its conclusion was that schema markup has no measurable effect on LLM visibility across those three systems. The author also noted that the schema distribution was driven in part by uneven sample sizes across domains. The page did not state how many domains or citations were analyzed.
Read this as a useful negative signal, not a controlled experiment. It compares existing sites, so it cannot isolate cause, and the missing sample counts limit how far you can lean on it.
| Source | Claim about schema and AI answers | Strength of evidence |
|---|---|---|
| Google Search Central, AI features documentation | No special schema.org markup needed for AI Overviews or AI Mode | Official platform guidance |
| Search Atlas, December 14, 2025 | No measurable effect on LLM visibility across OpenAI, Gemini, Perplexity | Observational, vendor-run, sample counts not published |
| Vendor sales material | Schema increases AI citations | Usually correlation or anecdote |
Where schema still earns its place
Entity clarity: how unambiguously a page tells machines who or what it is about. Organization, Article, Product, and FAQ markup state those facts in a consistent format, which supports classic search features and helps validators and crawlers that do read structured data.
The benefit is real but modest and indirect. Markup that disagrees with visible text can be ignored or treated as a quality problem, so accuracy matters more than volume. For a small business, a correct Organization and Article block on key pages is reasonable maintenance, priced as maintenance.
None of that changes the larger point. If language models extract mostly from visible page content, as the passage-level approach in our piece on what AI search means for SEO assumes, the page’s text is the asset to improve first.
How to vet a schema proposal
A schema line item is easy to scope and easy to over-sell. These checks keep it honest, and they pair with our guidance on choosing an AI marketing agency.
- Ask what outcome the markup targets: rich results, entity clarity, or AI citations.
- Ask for the platform documentation that says the markup is read for that outcome.
- Confirm the markup mirrors visible page text and validates without errors.
- Request a baseline of AI citations before work starts, using repeated prompts, as described in measuring AI search visibility.
- Check what else changes during the engagement, so schema does not get credit for new content.
What the evidence doesn’t support
Absence of evidence is not proof of absence. The Search Atlas analysis covers three AI systems at one point in time, and platforms can change what they parse. Google’s guidance concerns its own AI features and does not speak for ChatGPT or Perplexity.
We also looked for statements from Microsoft’s Bing team on schema and its language models, and for correlation figures from other studies circulating in secondary coverage. We could not confirm them from a primary source within our research budget, so we left them out. If a vendor cites one, ask for the original.
Frequently Asked Questions
Does schema markup help my site appear in Google AI Overviews?
Google says no special schema is required. Pages need to be indexed and eligible to appear in Search with a snippet, like any other result.
Should I remove schema I already have?
No. Accurate markup supports search features and entity clarity at low cost. Keep it correct, and stop expecting it to drive AI citations by itself.
What should I invest in instead for AI visibility?
Start with crawl access, indexable pages, and clear standalone passages that state a claim with its source. Then measure citations over repeated runs.
Sorting tactics that are documented from tactics that are merely sold is most of the job in AI search. At Grow with Hafsa we help teams build AI marketing systems where each tactic has a stated purpose, a baseline, and a result you can check. See also which AI bots to allow in robots.txt.