Use case
Structured data is how a search engine and an answer engine read a product, and it is capped by what your catalogue actually records. The audit tells you which missing attributes are costing you the most visibility, ranked, before anybody writes a line of markup.
Schema coverage cannot exceed attribute coverage. A structured data project ends at the second bar no matter how it is resourced.
Where you are
66%
Machine readability, product pages
the lowest of any page type measured
75%
Homepages, for comparison
the page with no distinguishing detail scores higher
74%
Category pages
also above the page that decides the purchase
US retail, measured by Adobe. The page holding every fact that separates one product from another is the worst structured page on the site. Adobe Analytics
The usual SEO remedy for this is a markup task. Add Product schema, add offers, add aggregateRating, validate it, ship it. That work is correct and it finishes quickly, because it can only express fields you already have.
Then the report comes back green and discovery has not moved, because the constraint was never the markup. Adobe scored US retail product pages at 66% machine readability, the lowest of any page type they measured, below homepages at 75% and category pages at 74%. The page carrying every fact that distinguishes one product from another is the worst structured page on the site.
The reason is that most of the distinguishing detail was written as prose, because prose is what sells. A richer description makes the gap wider rather than narrower: more has been said, and none of the new part is in a form a machine can compare.
Schema coverage is capped by attribute coverage. Markup work finishes fast and then stops, because there is nothing left to express.
Where the ceiling actually is
Not because the team ran out of time. Because there was nothing left to express.
An answer engine, asked about one of your products
1/3of the three are answerable, so the agent describes the product and then cannot recommend it. Search tolerates that ambiguity; an agent applies the constraint and drops you.
What it looks like
Marked up, recorded but not marked up, and never recorded at all. Only the last one needs new data.
Only the red band needs new data. The blue band is a markup task on facts you already own.
What happens
Three steps, and the markup is the last one rather than the first.
Which attributes a purchase in your category actually turns on, ranked by how much recording them would change what a buyer can decide. Not which recommended fields are empty: which absent fields are expensive.
Most of it is in your own titles and descriptions. On the shelf we measured, 161 of 216 missing values were recoverable from text the merchant already published. That is extraction, and it is the cheap half.
Once a fact is a value rather than a sentence, schema can carry it, a filter can use it and an agent can evaluate it. Same fact, three consumers, and none of them could reach it before.
What you get
Search engines and answer engines fail on the same missing field, for different reasons.
An index can infer from text and rank you anyway. An agent applying a constraint cannot: a missing value is treated as not qualifying, because that is the safe behaviour for it. The same gap costs you a position in one and the whole comparison in the other.
Adobe report AI-driven traffic to US retail sites up 393% year on year in Q1 2026, and converting 42% better than non-AI traffic, reversed from 38% worse a year earlier. Salesforce put AI agents behind more than 20% of holiday online retail sales.
Platform warnings tell you which recommended attributes are empty. They cannot tell you which of those empty attributes buyers in your category decide on, and those are different lists. Ours is ranked by the second.
The attribute that unblocks a filter on your own category page is usually the attribute that unblocks a shopping feed match and an agent comparison. You are not doing three projects.
Before you commit
Questions
One category export, and a ranked list of the attributes that are capping what a search engine and an answer engine can do with your products.
Request a Decision AuditNo code, no tag, no change to your site. The markup work comes after, and it is yours.