Will AI engines actually use your schema?

Why check crawler access in a schema validator?

Because markup on a page a crawler cannot fetch does not exist. Every other free validator grades the JSON-LD you hand it and stops there, which quietly assumes the hardest part is already solved. It often isn’t: a meaningful share of sites are blocking a retrieval crawler by accident, usually because someone pasted an “AI blocklist” from a blog post without knowing that GPTBot and OAI-SearchBot do completely different jobs.

Paste a URL rather than raw code and this tool reads your robots.txt crawler by crawler alongside the markup. Perfect Article schema on a page that disallows OAI-SearchBot scores badly here, and it should — no amount of markup quality can fix a page an engine never fetches.

Which crawlers actually produce citations?

AI crawlers do three different jobs, and conflating them is the most expensive mistake in this field. Training crawlers collect model training data — blocking them costs you nothing in visibility. Retrieval crawlers build the index an answer engine searches. Live-fetch crawlers pull a page the moment a real user’s question needs it. Block one of the last two and you disappear from that engine’s answers.

Does schema markup help you get cited by AI?

Probably not, and we would rather say so than sell you something. The best-controlled study available — Ahrefs, comparing 1,885 pages that added JSON-LD against roughly 4,000 controls — found no lift from schema, and a small decline in AI Overviews appearances. Separate experiments indicate answer engines extract the visible HTML during retrieval and largely ignore the JSON-LD block entirely.

What schema demonstrably does do is narrower and still worth the effort. It produces rich results in Google, which is a click-through lever regardless of AI. It disambiguates entities: sameAs pointing at Wikidata, Wikipedia or LinkedIn is the one schema property with a controlled positive result behind it, because knowledge-graph reconciliation is a real mechanism rather than a correlation. And it is what agentic interfaces read when they need your prices or your opening hours. That is the honest case, and this validator is built around it.

What does this tool check that others don’t?

Roughly two thirds of the 30 rules here overlap with any decent validator: syntax, required properties per type, absolute URLs, ISO dates. The rest are checks we added because they catch real, silent failures nobody else reports.

• Crawler access. Reads robots.txt per bot and tells you which answer engines are locked out.
• Client-side schema. If your JSON-LD is injected by Tag Manager, it is invisible to every AI crawler. We check the server HTML, not the rendered DOM.
• Contradictory dates. A dateModified earlier than datePublished, or a date in the future — both common CMS bugs, both unreported by every other tool we tested.
• Self-serving ratings. An aggregate rating attached to your own Organization is an explicit Google policy violation and a common cause of a manual action.
• Markup that contradicts the page. A price or headline in your schema that appears nowhere in the visible text.

Just as importantly, there are things we deliberately don’t flag.http://schema.org is valid. Google’s image minimum is 696px, not the 1200px widely repeated. Rating scales are declared by bestRating, so a 9.2/10 is legal. And an honest three-year-old dateModified is correct, not a fault — penalising it would only encourage cosmetic date-bumping, which is exactly what search engines are trying to detect.

Frequently asked questions