Schema markup for AI search: what we can and cannot show it changes

72 of 121 audited sites passed the Organization schema check. That leaves 47 sites where a basic business identity signal failed.

Schema markup can give AI search systems clearer labels for an organisation, product or answer. Our data does not test whether adding schema increases citations, so treat it as a clarity and consistency check rather than a citation switch.

The figures come from our free AI readiness check, which covers 121 audited sites. This is a sample of 121 audited sites, not a random sample of the whole web.

What the main schema blocks can tell an engine

Schema is a machine-readable layer around facts that need to exist on the page. It gives those facts labels and relationships, which can reduce the work an engine has to do when it decides what a company is, what it sells and which questions its page answers.

Organization schema and entity recognition

An Organization block helps establish that a named organisation is the entity behind a site. That matters when a founder uses a short brand name, a legal company name and a product name across different pages.

It is a signal, not a registration form. Adding it does not put a business into an AI assistant's catalogue, and it does not settle every conflicting fact found elsewhere.

The 47 failures make accuracy more important than simply adding a block. The organisation name, site URL and visible business details should describe the same company.

Product schema and software identity

Product or SoftwareApplication schema gives an engine a clearer object to identify. The object might be a physical product, a software product or an application with its own name and offer.

Only 60 of the 121 audited sites passed this check. That leaves 59 failures. A software page may still have a title, a price and a product description, but those are not the same as a labelled product entity.

Keep the product facts aligned. If the page describes one product and the block describes another, the extra structure creates more ambiguity, not less.

FAQ schema and answer recognition

An FAQ block connects a question with the answer beneath it. This can make a page's scope easier to parse when buyers ask about a specific feature, use case or limitation.

The relevant readiness check passed on 56 of 121 sites and warned on 63. Treat it as a content and structure test, not a promise of a special answer result.

A block cannot supply an answer that the page does not contain. It can label a clear answer, but it cannot turn a vague sales line into useful evidence.

The pass rates show where sites stop

The schema checks sit among other basic signals in the same sample. The wider pattern matters because markup works alongside access, clarity and visible content.

The readiness results show that some foundations are already common. A clear H1 passed on 101 of 121 sites. A sitemap passed on 100. Content was present without JavaScript on 92.

Other signals were much weaker. A plain statement of what a site is, who it is for, what it costs and what it is against passed on only 20 of 121 sites. Comparison or alternatives pages passed on 54. Questions answered on the page passed on 56.

That gives you an order of work. Make the business and product understandable in visible text first. Then use schema to label the same facts.

A validator can tell you whether the JSON is syntactically valid. It cannot tell you whether the page makes a defensible statement about the business.

Access still comes first. AI crawlers were allowed in robots.txt on 116 of 121 sites. If a crawler cannot fetch the page, a well-formed block is sitting behind a locked door.

The check for the site's own pages getting cited passed on only 5 of 121 sites. That is a different outcome from passing Organization or Product schema. Clean structured data and observed citations are separate measurements.

Schema is worth checking, not assuming will work. Schema can pass validation cleanly while the page still fails to convince a buyer or an engine. The two checks measure different things.

A schema markup example with the numbers visible

Take a typical SaaS pricing page from the 121-site sample. The page has a company name, a product name, a price and buyer questions. Walk through the checks in this order.

1. Check the organisation. The Organization check passed on 72 of 121 sites and failed on 47. For this example page, confirm that the block names the business behind the product, points to the correct site and does not use an old brand name.

2. Connect the product. The Product or SoftwareApplication check passed on 60 of 121 sites and failed on 59. On the example page, confirm that the product name in the block matches the product named on the page, rather than naming only the parent company.

3. Check the buyer facts. The price visible or pricing page linked check passed on 90 of 121 sites. On the example page, compare the offer and price in the structured data with what a buyer can see or reach from the page.

4. Add answer structure where it earns its place. The questions answered on the page check passed on 56 of 121 sites and warned on 63. On the example page, add an FAQ block only for questions with visible answers, such as who the product is for or how its pricing works.

5. Test the result as a reader and a machine. Run a schema validator, then read the rendered example page without relying on the code. The technical test catches broken structure. The page review catches claims that are missing, stale or hard to understand.

The important part is the relationship between the blocks. Organization says who is behind the site. Product says what the named offer is. FAQ says which questions the page answers. They should describe the same business and offer.

JSON-LD matters here for a simpler reason: it is a format, not a guarantee. It carries labels. The page still has to carry meaning.

A schema markup generator can speed up a first draft. It may also produce fields you do not need or cannot support. Review every generated value against the visible page before publishing it.

What markup cannot prove

Schema can help an engine read an entity. It cannot show that the engine will choose that entity for a recommendation. Those are separate steps.

The citation data shows the setting in which that choice happens. In the weekly research on who owns AI answers, listicles and roundups account for 67.5% of recorded citations, competitor product pages account for 24.4%, and the product's own site accounts for 1.5%.

This is descriptive scan data, not a controlled test of schema. We have not measured a before-and-after change that isolates markup, so we cannot attribute citation rates to an Organization, Product or FAQ block.

A failed schema check does not guarantee poor AI visibility either. The same site may be named on a comparison page, included in a roundup or discussed in a community thread. Community threads account for 3.4% of citations, documentation for 1.1% and social posts for 0.8% in the scan data.

Measure the two outcomes separately. First, check whether the site states and structures its identity clearly. Second, check whether assistants actually name the business or cite its pages for the buyer questions that matter.

The free AI visibility checker is for the second kind of check, while the readiness sample is for the first. The opening 72 of 121 result tells you where Organization markup stands in this sample. It does not tell you that the block caused, or will cause, a citation.