What does your page’s region list say?

The list nobody who built the page has ever seen

No browser shows it to you. It is not in the rendered page, it is not in view-source in any readable form, and it does not appear in a screenshot. It is assembled by the browser from your tags and your role attributes, handed to assistive technology, and opened by a keystroke you have probably never pressed. So it is the one part of a site that ships unreviewed by definition — not because anyone was careless, but because there was nowhere to look.

Here is what it reads on three sites that are visibly trying, read live on 22 September 2026:

nasa.gov        banner · banner · search · Primary navigation, navigation ·
                search · main · contentinfo
                → two unnamed banners, two unnamed searches

developer.mozilla.org
                banner · navigation · navigation · contentinfo
                → two unnamed navigations, and no main region at all

gov.uk          Cookies on GOV.UK, region · banner ·
                (labelled by #super-navigation-menu-heading), navigation ·
                Site-wide, search · main · search · contentinfo
                → named, distinct, and the one that reads cleanest

None of those is a conformance failure, and NASA’s two banners are a mobile header and a desktop header that are each correct on their own. That is the point: this is a class of defect that survives every review a serious team runs, because the artefact it lives in is one nobody is shown.

What WCAG actually says about landmarks

Less than most checkers imply. No success criterion requires a page to have a <main>, or a banner, or any landmark at all. SC 1.3.1 Info and Relationships asks that structure conveyed visually be available programmatically, and headings alone can satisfy it. SC 2.4.1 Bypass Blocks asks for a way past repeated blocks, and lists landmarks (technique ARIA11) as one sufficient way — with a skip link as another, which is why a page with a working skip link and no landmarks conforms.

So this checker calls exactly three things failures, and all three are markup contradicting itself: two elements both claiming to be the main region, a label pointing at an element that does not exist, and a landmark hidden from the technology it exists to serve. Everything else is reported for a look, with the criterion named where one genuinely applies and left off where it does not.

That restraint is the tool. Telling a stranger their working site “fails WCAG” because it has no <aside> is how accessibility scanners lose the room.

What this checker refuses to call a failure

Landmarks are unusual: the defect and the correct shape are often the same tag, and what separates them is how deeply it is nested. So the negatives are pinned as hard as the positives, by scripts/check-landmark-audit.mjs in the build:

<!-- correct — scoped to the article, NOT a second banner -->
<main><article><header>Post title</header></article></main>

<!-- correct — generic container, exactly like a div -->
<section>…</section>

<!-- correct — redundant, harmless, never reported -->
<nav role="navigation" aria-label="Primary">…</nav>

<!-- correct — this IS the fix for duplicate navs -->
<nav aria-label="Primary">…</nav>   <nav aria-label="Footer">…</nav>

<!-- fails 1.3.1 — two main regions, ambiguous jump target -->
<main>…</main>   <div role="main">…</div>

<!-- fails 1.3.1 — ARIA drops the name silently, no fallback -->
<nav aria-labelledby="nav-heading">…</nav>   <!-- #nav-heading doesn't exist -->

<!-- the common one — the list reads "navigation, navigation" -->
<header><nav>…</nav></header>   <footer><nav>…</nav></footer>

A <header>, <footer> or <aside> inside <article>, <aside>, <main>, <nav> or <section> exposes no role at all, per the HTML accessibility mapping — so it is never counted, never reported, and never called a duplicate. Every blog post on the internet has a nested <header>, which is what makes that the loudest false positive available here.

And the honest gap: we read the HTML your server sends and do not run your JavaScript. A page whose markup has no landmarks and almost no content is reported as verify — a client-rendered shell — rather than as a page with an empty region list.

The fix, in full

Four tag names and two attributes, all of them in the layout template, which is why fixing it once fixes every page.

<body>
  <header>                              <!-- banner -->
    <nav aria-label="Primary">…</nav>   <!-- navigation, "Primary" -->
  </header>

  <main>…</main>                        <!-- main — exactly one -->

  <aside aria-label="Related">…</aside> <!-- complementary -->

  <footer>                              <!-- contentinfo -->
    <nav aria-label="Footer">…</nav>    <!-- navigation, "Footer" -->
  </footer>
</body>

That is a region list of six entries, every one of them distinguishable, and it costs two aria-label attributes over the markup most sites already have. The two words in those labels are the entire difference between a shortcut and a list the reader has to open one by one.

Then resist adding more. Past about eight regions the list takes as long to read as the page, which is the failure mode of every well-intentioned landmark cleanup: a dozen named <section> elements that turn a shortcut back into a scan.

And check the other half of SC 2.4.1 while you are in there. If your page has no main region, the skip link is the only bypass mechanism you have left — the skip link checker follows its href and tells you whether it still points at anything.

Common questions

What is a landmark, in one sentence?

A region of the page the browser hands to assistive technology as a named destination — banner, navigation, main, complementary, contentinfo, region, form and search. A screen reader user presses one key and gets the list of them, then jumps. It is the equivalent of glancing at a page and going straight to the bit you want, and it is the primary way somebody moves around a site they have not seen before.

Does WCAG require landmarks?

No, and any tool that tells you otherwise is overselling. SC 1.3.1 Info and Relationships asks that structure conveyed visually be available programmatically, which headings alone can satisfy. SC 2.4.1 Bypass Blocks asks for a way past repeated blocks, and lists landmarks (technique ARIA11) as one sufficient way and a skip link as another — so a page with a working skip link and no landmarks at all conforms. That is exactly why this checker calls only three things failures, and all three are markup contradicting itself.

So what does it actually call a failure?

Three things. Two elements both claiming to be the main region, which makes the jump-to-main command ambiguous. An aria-labelledby pointing at an id that does not exist, where ARIA silently drops the name rather than falling back — so the markup looks labelled forever. And aria-hidden="true" on a landmark, which deletes the region and everything inside it from the accessibility tree while leaving it on screen. Each one is a structure that was declared on purpose and then declared wrongly, which is what 1.3.1 is for.

Why does "navigation, navigation" matter?

Because the region list is a shortcut, and a shortcut with identical entries is not one. A page with a header menu, a sidebar and a footer sitemap gives a reader three entries that all say "navigation", so they have to open each one to find out which is which — which is more work than tabbing, the thing the list existed to avoid. Two aria-labels fix it: aria-label="Primary" and aria-label="Footer". This is the most common landmark defect on the web.

My <header> is inside <article>. Isn't that a banner?

No, and this is the trap that makes naive landmark checkers useless. Per the HTML accessibility mapping, a <header> or <footer> inside <article>, <aside>, <main>, <nav> or <section> is scoped to that section and exposes no role at all — it is not a banner and not contentinfo. The same rule applies to <aside>, which is only complementary at the top level. Every blog post on the internet has a nested <header>, so a checker that counted them would report duplicate banners everywhere. This one tracks nesting and does not.

Are my <section> elements regions?

Only the ones with an accessible name. A <section> with no aria-label and no resolving aria-labelledby is exposed as a generic container, exactly like a div — that is the spec, not an oversight, and it is not a defect. It is worth saying because teams routinely ship a page of nameless sections expecting a navigable outline that no screen reader will ever show them. The same rule applies to <form>: nameless, and it is not a form landmark.

Is <nav role="navigation"> wrong?

It is redundant and completely harmless — the browser maps the same role either way — and this checker never reports it. The first rule of ARIA is not to use it where HTML already does the job, so it is worth removing on a tidy-up, but it breaks nothing and it is not what you came here for.

How many landmarks should a page have?

Few enough that the list is still a shortcut. Banner, main, contentinfo and one or two navigations covers most pages. Once you are past about eight, scanning the region list costs as much as scanning the page, and the reader stops using it. If you are adding names to a dozen <section> elements to make them regions, that is the direction to stop going in.

Why did you say my page has no landmarks when it clearly does?

Almost always because they arrive with your JavaScript. This reads the HTML your server sends, deliberately — it is fast, free and the same markup a crawler sees. When that HTML has no landmarks AND almost no content, the result is reported as verify rather than as a finding, and says so. The free WCAG 2.1 AA scan renders each page in a real browser and sees what mounts.

Can you tell me if my landmark names are good?

Only the narrow case where a name is the role word itself — aria-label="Navigation" on a <nav>, announced as "navigation navigation". Beyond that, no: "Section 2" is a name as far as any machine is concerned. Naming a region by what it contains rather than what it is remains a human decision, and it takes about four seconds per landmark.

Does this store my domain?

No. The check is an HTTP GET of the page you name, run inside the request and discarded when it answers. There is no account, no email field and no row written anywhere. If you go on to run the free scan, that one does keep a report so you can come back to it, which is the whole point of a report.