7 Schema Markup Mistakes That Quietly Kill Your AI Visibility
Structured data is the cheapest AEO win there is: it tells an answer engine what your page means instead of making it guess. But most schema in the wild is either broken, boilerplate, or quietly contradicting the visible page — and a contradicted claim is worse than no claim at all. Here are the mistakes we find most often, and what to do instead.
1. Schema that disagrees with the page
The fastest way to get your markup discounted is to assert something the page itself does not show. Prices in JSON-LD that differ from the pricing table. Hours that say open Sunday when the footer says closed. A review count of 480 with three reviews on screen. Engines cross-check the machine layer against the human layer; when they disagree, the machine layer loses and your whole entity gets a lower trust weight.
Fix: generate schema from the same source of truth that renders the page, never by hand-editing a snippet someone pasted in 2023.
2. No entity spine (@id, sameAs, and one canonical name)
Individual pages markup up nicely, and yet nothing connects them. Without stable @id values and sameAs links to your profiles, each page reads as an unrelated fragment instead of evidence about one business. Answer engines recommend entities, not orphan pages.
- One
OrganizationorLocalBusinessnode with a permanent@id, referenced by every other node. sameAspointing to your main profiles and listings — the same ones the engines already cite (see which sources AI trusts).- Exactly one spelling of your name, address and phone everywhere. No “LLC” on one page and not the next.
3. The wrong type, or a type stuffed with everything
A services page marked as Product. A blog post marked as NewsArticle. A single node claiming to be an Organization, a Service, a Product and a FAQ at once. Over-claiming does not multiply your chances; it makes the record ambiguous and easy to drop.
The mistakes and why they cost you
| Mistake | Why it hurts AI visibility |
|---|---|
| Markup contradicts visible content | Machine layer is discounted; entity trust drops sitewide |
Missing @id / sameAs | Pages never resolve into one recommendable entity |
Self-serving aggregateRating | Ignored, and flags the source as promotional |
| FAQ schema with no answer text on the page | Nothing quotable; the citation goes to a competitor |
| Schema injected only by client-side JavaScript | Many AI crawlers never execute it — effectively invisible |
| Copy-paste boilerplate across every page | No page-level specificity, so no page wins a specific question |
| Stale dates, prices, staff, locations | Engines prefer fresh, verifiable sources over confident old ones |
4. Marking up answers you never actually wrote
FAQ and HowTo markup only helps if the answer exists in plain text on the page, in a form a model can lift in one or two sentences. Schema is a label, not a substitute for content. Write the direct answer first — short, specific, with the number or the condition in it — then wrap it.
5. Delivery problems that make perfect schema moot
We regularly audit sites with flawless JSON-LD that no answer engine can read: the markup renders client-side, the page is blocked to AI user agents, or the canonical points somewhere else so the markup is attributed to a URL nobody cites. Check delivery before you check syntax.
Do this in order
- Confirm AI crawlers are allowed and your schema is in the server-rendered HTML.
- Publish one canonical entity node; reference it from every page.
- Match every asserted value to something visible on the page.
- Write the quotable answers, then mark them up — not the reverse.
- Re-check after every template or CMS change; schema silently breaks on redesigns.
The deeper mechanics live in our guides to schema markup for AI, what AEO is, how to do AEO, and the AEO checklist.
Find out what the engines currently say about you
Schema is a means to an end: being named when a buyer asks. The free Rooster AI Visibility Scan shows which answer engines mention your business today, which sources they cite instead, and where your structured data is letting you down — and the AI Visibility Tracker watches it week over week as you fix things.