Your schema is lying to AI — and AI can check

When we verified the schema claim by claim, the markup fell apart — not because it was fake, but because it was inconsistent with every other source describing the same business.

The images were broken. The schema's logo and image URLs pointed to /wp-content/uploads/ paths — WordPress paths, on a site that had migrated to a different platform. The schema had been written for the old site and carried over, unchecked. Every engine following those links gets a 404.

The opening hours disagreed with Google. The schema said open until 8pm, seven days a week. Google Business Profile said closing at 5pm. Ask an AI assistant "is this shop open right now?" and it gives a wrong answer no matter which source it believes.

The map coordinates were off. The schema's geo coordinates and the site's own navigation link pointed to two different spots in the city.

The social profiles didn't match. The schema listed a LinkedIn company page; the site footer linked a LinkedIn personal profile. Three profiles that appeared in the footer were missing from the schema entirely. One linked profile didn't appear to exist at all.

And one validation error meant the whole block risked being ignored by parsers anyway.

Here's the thing most guides to structured data won't tell you: a missing schema block is a neutral signal. An inconsistent one is a negative signal.

AI engines don't read your schema in isolation. They cross-reference it — against your visible pages, your Google Business Profile, your directory listings, your map presence. That cross-referencing is precisely how they decide whether you're one real, reliable business or a pile of half-maintained claims.

Every contradiction is a small answer to the question "can I trust what this site tells me?" — and the answer is no. A broken image link is a claim that failed verification. Mismatched hours are two sources calling each other liars. A dead social link is a reference that doesn't check out.

The engines never have to decide which version is true. They just quietly recommend someone whose story checks out instead.

That's how a business with a thousand five-star reviews loses every AI recommendation to competitors with weaker reputations and tidier data.

You don't need an agency to find out if your schema is lying. You need a browser:

1. Open your homepage, view the page source (type view-source: before your URL), and search for ld+json.
2. No results? You have no structured data. That's a gap, but a clean one — you can build it right.
3. Found a block? Paste it into validator.schema.org. Errors are strike one.
4. Now open every URL in it. The logo. The images. Anything that 404s is a broken claim.
5. Finally, put your schema next to your Google Business Profile and compare line by line: name, address, phone, hours, coordinates. Any disagreement is two of your own sources contradicting each other — and machines noticing.

Most businesses that run this check find at least one contradiction.

Structured data isn't a box you tick once. It's a promise you keep — across every platform migration, every change of opening hours, every new social account. The businesses AI engines recommend aren't the ones with the most markup. They're the ones whose story is identical everywhere a machine looks.

The Happy Cat runs free AI visibility audits for businesses that want to know what ChatGPT, Perplexity and Gemini say about them — and why. Request yours here.