Schema audit
Whether a machine reading your pages can tell what they are, who publishes them and what you sell — and the specific fixes when it cannot.
An AI system reading your page has to work out what it is looking at. Structured data is how you tell it directly instead of leaving it to inference. The Schema audit checks whether you do, page by page.
What is checked
Identity. Does the page declare the organisation or person behind it, with a consistent identifier across the site? Sites that split their identity across variants of their own domain end up read as two different entities.
Page type. Is a product page marked as a product, an article as an article, a listing as a listing? Checks are page-type aware — an editorial signal is not demanded of a category page, and a missing breadcrumb is not a finding on the homepage.
Commerce details, on pages that sell: price, availability, currency, condition, and whether review markup is complete enough to be usable.
Metadata and previews. Title, meta description, canonical URL, language declaration, hreflang for multi-language sites, and social preview tags.
Consistency. Whether the declared language matches the page's own alternates, whether the canonical points at the right host, and whether declared profiles actually correspond to the entity on the page.
Reading the results
You get a site score, a list of audited pages with individual scores, and per-page findings. Work from the page list — the site score tracks progress, individual pages hold the work.
Each finding states what is wrong on that page, what to change, and what the page would look like afterwards. Findings describe the page in front of them rather than a generic template: if a fix says a block is missing, it is missing there.
Findings that do not apply to a page's type are marked as such instead of being counted against it.
Prioritising
- Identity first. Everything else attaches to a page whose subject is clear.
- Then page type. A product page not identifiable as a product is invisible to anything asking about products.
- Then commerce and metadata detail. Incomplete markup is often ignored wholesale rather than partially used, so finishing a block matters more than starting another.
- Re-run after deploying. Results persist, so you can compare against the previous state.
What this audit does not tell you
It measures whether your page is legible, not whether it is reachable or convincing. A perfectly marked-up page that AI crawlers are blocked from, or that renders empty without JavaScript, gains nothing:
- Reachability → Robots.txt audit
- Readability without JavaScript → Technical checks
- Whether it is actually cited → Sources tab
Related
- Report tab — where the audits live
- Technical checks — the layer beneath this one
Updated 2026-08-03