Schema markup that validates perfectly can still hurt AI visibility, and five specific mistakes explain why, according to Search Engine Journal's latest Ask An SEO column by Helen Pollitt. The core issue: large language models are not just parsing page types — they are looking for reassurance of entities and the relationships between them.

The question posed to the column was direct: "What are the most common structured data mistakes that hurt AI visibility?" Pollitt's answer reframes schema from a technical checklist into an entity strategy problem, with concrete failure modes that span manual penalty risk and contradicting machine-readable facts.

Why isn't a schema checklist enough for LLMs?

Many SEOs mark up the main schema types for a page as a matter of rote, Pollitt writes. That approach worked when the goal was simply helping search bots classify content. AI search raises the bar.

"The key for AI search is realizing that the LLMs are looking for reassurance of entities and relationships," the column states. Organizations need to ask: "Does our structured data make it easier for machines to understand the context and relationship of information on our page?"

Her example: an ecommerce blog article using Article schema is technically correct. It tells bots the content is informational, not commercial. But it says nothing about how that article connects to other entities on the site. A stronger implementation links the article to its author's page and to the organization the author works for — layering Article, Person and Organization schema together to give machines cross-site context.

Can your own markup outvote the rest of the web?

No — and that is the second mistake. Pollitt warns that marking up correct information on your own site is not enough, because it is not a given that LLMs will trust your markup as the most authoritative version.

"It is not a given that your marked up content will be the information the LLMs trust as the most authoritative," she writes. If third-party sites still carry your old brand name or outdated product pricing, your correct on-site schema may lose the credibility contest. Part of a structured data strategy for AI visibility, she argues, is actively correcting misinformation elsewhere on the web.

What do consistent entity identifiers fix?

The third mistake is failing to use @id references. Defining an entity once — for example, an Organization node at "https://www.helenseocommercestore.example/#organization" — and referencing that @id in subsequent markup keeps entity data consistent across every page and template without duplicating code.

Pollitt illustrates the cost of skipping this with a dated scenario: organization schema added to product pages in 2022 under the old name "Helen's Shop," while the homepage schema was updated in 2026 to "Helen's Ecommerce Store." The conflicting names force AI agents to guess which is correct. With a shared @id, the 2026 correction would have propagated automatically to every page referencing that entity — no conflicts, no repeated updates.

When does valid schema become a penalty risk?

The fourth mistake is markup that does not match visible content — a risk for both traditional SEO and LLM optimization. Pollitt's example is a coffee machine product page showing a £79.99 price and no customer reviews, while its Product schema declares an aggregateRating of 4.9 with a reviewCount of 127.

That gap carries real consequences. Intentionally marking up content that does not exist on the page, she writes, "risks a manual penalty from Google" and is "just plain confusing for LLMs," because the signals in schema need reinforcement from the on-page content itself.

What happens when schema and page disagree on facts?

The fifth mistake is stale or contradicting markup. In a second version of the same product page, the visible price is £79.99 but the Offer schema records 59.99 GBP — a £20 discrepancy.

"AI systems are trying to establish fact," Pollitt writes. "When the two 'statements' of fact are directly contradicting each other, this reduces the reliability of the webpage as a source of information about the product." At best this confuses the bots; at worst it damages customer satisfaction and invites a manual penalty.

What to monitor next

For sites optimizing for AI search, the actionable audit points are clear: entity relationships between articles, authors and organizations; external data consistency for brand names and pricing; @id usage to prevent cross-page conflicts; and schema-to-content parity on every product page. As LLM citations increasingly depend on which sources the models deem trustworthy, sites that resolve these five conflicts now are better positioned when the next wave of AI search systems evaluates them.