Structure & credibility

See the technical signals AI systems use to judge your page

Schema validity, page-type classification, heading structure, metadata, accessibility markup. The things you can control and check — separate from your content, and separate from your backlinks. You get a grade, the evidence behind every check, and the fix.

blog.acme.com/ai-guide
{ } Article schema · JSON-LD
✓ Schema found
How to Get Cited by AI in 2025
✓ Clear H1
Jane Smith · Acme Research·Updated Apr 2025
✓ Author entity
🖼 alt=""
⚠ Alt text missing
Why it matters

Credibility a person reads instantly is invisible in the markup.

An assistant deciding whether to cite you has none of the cues you rely on. It has the HTML.

What a person sees

A polished page from a company they have heard of, with a photo of the author, a date near the top and prices in a clean table. They trust it in about a second.

What a machine can prove

Whether an Article or Product type is declared, whether the author is a named entity rather than a picture, whether the date is a machine-readable field, and whether the price is a value or a rendered image.

What gets checked

Every check, with a verdict and a reason to care

Checks are selected for the kind of page you gave us — a pricing page is not graded like an article — then grouped by how much each one actually changes.

61

Authority score

Three required checks are failing. Fixing those two schema issues moves the score more than everything else on the page combined.

Passed14
Warnings4
Failed3

Required

critical

Missing any of these is why a page does not get quoted.

No Product schema on a page selling a productNothing declares what this page is about, so nothing can index it as an offer. JSON-LD
Price appears in text but not as a structured valueA number a person can read and a machine cannot use. JSON-LD
Single, unambiguous H1 headings
Canonical URL declared and self-referencing meta

Important

major

These decide whether you are the source or a supporting link.

No named author or publisher entityThe claims on the page belong to nobody a machine can identify. JSON-LD
! Last modified date is over 14 months oldFreshness is a tie-breaker when two pages answer equally well. meta
Headings describe content rather than styling it headings

Supporting

advisory

Worth doing, but not what is holding you back today.

! No FAQPage markup despite a question-and-answer section JSON-LD
Images carry descriptive alt text markup
Priority Each check's priority comes from its own severity, not from where it happens to appear in the report — critical reads as Required, major as Important, advisory as Supporting, in every surface including the PDF.

Find out what your markup actually says.

Run the check on any URL. You get every check with its verdict, ordered so the three that matter are at the top.

FAQ

Structure, answered

Is this backlinks and domain authority?

No. Nothing here looks at other sites, your domain’s reputation, or anything a traditional SEO tool scores. It reads the signals on the page itself — the ones that let a machine work out what this page is, who published it and whether it can be attributed. A brand-new domain can pass every check here.

So is it a schema validator?

Schema is one input. The check also reads page type, canonical and metadata, heading structure, publication and update dates, identity signals, and the accessible names on the page. A page with flawless JSON-LD and no discernible structure still fails, and a page with no JSON-LD at all can pass the checks that matter most for it.

What does page-type classification change?

What is expected of the page. An article should have an author and a date; a product page should not be judged for lacking them. We classify the page first, then apply the checks that belong to that type — which is why a wrong classification distorts everything after it, and why it is worth looking at first when the results seem odd.

Why does heading structure matter to a machine?

Headings are how a page says where one idea stops and the next begins. A system pulling one passage out of your page uses them to decide where to cut. Where every heading is styled text with no structure behind it, the page is one undifferentiated block and the passage that gets quoted is whatever happened to be nearby.

Why is accessibility markup in an AI visibility check?

Because the accessible name is the only name a machine has for a control or a region. It is the same information a screen reader uses, which means the work counts twice — and it is the reason a page that reads well often turns out to be operable too.

What is the difference between Required, Important and Supporting?

How much the check costs you when it fails. Required means something most credible pages have is missing. Important is a real weakness. Supporting strengthens a page that is already sound. Work down that order — a Supporting check passing on a page that fails a Required one changes nothing.

Which metadata actually counts?

The ones that establish identity and freshness: a canonical URL so a machine knows which version is authoritative, a title and description that describe this page rather than the site, and publication and update dates where the page is the kind that ages. Most of the rest is inherited from SEO practice and does very little here.

Do I need schema on every page?

No. Organisation markup once, sitewide, does most of the work. Beyond that, add the type that matches the page — Article on articles, Product on products — and skip it where nothing fits rather than forcing a shape onto a page that is not that shape.

What if my schema is generated by a plugin?

It is checked the same way, and plugin output is where a lot of findings come from. The usual pattern is markup that is technically valid and factually thin: an author field with no author, a dateModified that never moves, a type that was right for the template and wrong for this page.

Does any of this guarantee I get cited?

No. These are the signals that make a page usable and attributable. Whether it gets used also depends on what it says, how recent it is, and who else answered the same question — none of which is decided by markup.