More ARIA can make a page less accessible because ARIA changes what assistive technology is told about an element, and nothing checks that the claim is true. A role="button" on a <div> tells a screen reader "this is a button". It does not make the <div> focusable or respond to Enter. The W3C's Authoring Practices Guide calls this out under the heading "No ARIA is better than Bad ARIA" and warns that "incorrect ARIA misrepresents visual experiences, with potentially devastating effects" for screen reader users (W3C WAI, APG Read Me First). The data points the same way. In WebAIM's February 2026 scan of the top million home pages, pages with ARIA averaged 59.1 detected errors and pages without averaged 42 (WebAIM Million 2026). WebAIM is careful to say that doesn't prove ARIA caused the errors, since those pages are also more complex. Most ARIA accessibility problems come from a few repeat patterns: redundant roles, roles without behaviour, and aria-hidden on things people can still tab to. The fix is usually less ARIA, not more. Use the native HTML element and keep ARIA for widgets HTML doesn't have.
What does the WebAIM Million 2026 say about ARIA?
WebAIM runs the WAVE engine over the rendered DOM of the top million home pages each year. The February 2026 numbers on ARIA (WebAIM Million 2026):
- 82.7% of home pages used ARIA (not counting landmark roles), up from 79.4% in 2025.
- WAVE found 133,589,803 ARIA attributes, over 133 per page. That's up 27% in a year, and up from 22 per page in 2019.
- Pages averaged 23.3
aria-hidden="true"attributes, up 30% on 2025. - Of the 5.7% of pages with an ARIA menu, 22% of those menus introduced barriers.
- Pages with ARIA averaged 59.1 errors against 42 without, which WebAIM frames as about 17 more potential barriers per page.
WebAIM's caveat matters, so here it is in full context: "This does not necessarily mean that ARIA introduced these errors", because "these pages were also more complex" (WebAIM Million 2026). A page with a mega-menu, a carousel and three third-party embeds will have more ARIA and more errors whether or not the ARIA is to blame. I wrote up the wider report in what the WebAIM Million 2026 tells us.
So the honest reading is narrower than the headline: ARIA use is growing fast, and it isn't buying fewer errors.
Why does ARIA cause accessibility problems?
ARIA only changes the accessibility tree. It's a set of labels that browsers pass to assistive technology, and the APG describes it as "analogous to a CSS for assistive technologies" (W3C WAI, APG Read Me First). CSS changes how something looks without changing what it does. ARIA changes what something is announced as without changing what it does.
That's the whole problem. The APG says "a role is a promise", and spells out what the promise is: a <div role="button"> "is a promise that the author of that <div> has also incorporated JavaScript that provides the keyboard interactions expected for a button". It also notes that "ARIA roles do not cause browsers to provide keyboard behaviors or styling" (W3C WAI, APG Read Me First).
A native <button> keeps its promise for free. It's focusable, it fires on Enter and Space, and it reports its disabled state. A <div role="button"> keeps none of those promises until someone writes the code, and the screen reader user can't tell which one they've got until it fails.
ARIA can also hide things. The APG calls part of ARIA "like a cloak; it covers up, or overrides, the original semantics or content", and says "authors can inadvertently override accessibility semantics" (W3C WAI, APG Read Me First). An aria-label on a link replaces the visible link text for a screen reader. If the two drift apart after a copy edit, sighted and non-sighted users now see different pages.
What are the most common ARIA mistakes?
These are the patterns I see most often in scan results and code review. Each one maps to written W3C guidance.
Redundant roles. <button role="button">, <nav role="navigation">, <ul role="list">. The W3C's ARIA in HTML Recommendation says setting role or aria-* to values matching the element's implicit semantics is "NOT RECOMMENDED", because "doing so is unnecessary and can potentially lead to unintended consequences" (W3C, ARIA in HTML). It's mostly noise, and a sign the author wasn't sure what the element already did.
Overriding native semantics. The old "Using ARIA" note gives the classic example: <h2 role=tab>heading tab</h2> wipes out the heading, so it vanishes from a screen reader's headings list. The suggested fix wraps instead: <div role=tab><h2>heading tab</h2></div> (W3C, Using ARIA). ARIA in HTML makes the rule formal: authors "MUST NOT use the ARIA role and aria-* attributes in a manner that conflicts with the semantics" in its conformance tables (W3C, ARIA in HTML).
Roles without behaviour. role="checkbox" with no aria-checked, or role="tab" with no arrow-key handling. Using ARIA's third rule is "All interactive ARIA controls must be usable with the keyboard" (W3C, Using ARIA). WebAIM's figure that 22% of ARIA menus introduced barriers belongs here (WebAIM Million 2026). role="menu" promises application-style arrow-key navigation, and most site navigation is a list of links that never needed it.
aria-hidden on focusable content. The fourth rule is blunt: "Do not use role="presentation" or aria-hidden="true" on a focusable element", because it "will result in some users focusing on 'nothing'" (W3C, Using ARIA). The W3C ACT rule for this case fails exactly this markup (W3C ACT rule 6cfa84):
<div aria-hidden="true">
<a href="/" style="position:absolute; top:-999em">Link</a>
</div>
A keyboard user tabs onto that link and their screen reader announces nothing. The rule maps to WCAG 4.1.2 Name, Role, Value. With home pages averaging 23 aria-hidden attributes each, one of them landing on a modal's close button or an off-canvas menu is not unusual.
Labels that contradict the visible text. <a href="/pricing" aria-label="Learn more">See pricing</a>. A voice-control user says "click See pricing" and nothing happens, because the accessible name is "Learn more". WCAG 2.5.3 Label in Name exists for this case (WCAG 2.2).
How do I fix ARIA accessibility problems?
Mostly by deleting. Work through it in this order.
- Replace the element, not the attributes. If a
<div>hasrole="button",tabindex="0"and a keydown handler, swap it for<button type="button">and delete all three. The first rule of ARIA, still the best one-line summary, is that if a native HTML element "with the semantics and behavior you require already built in" exists, use it (W3C, Using ARIA). - Strip redundant roles. Search the codebase for
role="button"on<button>,role="link"on<a href>,role="navigation"on<nav>. Removing them changes nothing for users. - Audit every
aria-hidden="true". For each one, ask whether anything inside can take focus. If it can, either removearia-hiddenor make the content inert. The HTMLinertattribute handles both the hiding and the focus in one go. - Finish or remove custom widgets. A role you keep has to come with its states and keyboard model. The APG has a pattern for each widget (tabs, combobox, disclosure, menu button). If you don't have time to build the full pattern, use the simpler native control instead. A list of links is a better nav than a half-built
role="menu". - Make
aria-labelthe last resort. Prefer visible text, then<label for>, thenaria-labelledbypointing at visible text. Keeparia-labelfor icon-only controls. My Name, Role, Value guide covers the order in detail.
The test afterwards is simple: tab through the component, then do it again with a screen reader on. If you haven't done that before, start with how to test with a screen reader.
Can an accessibility scanner catch ARIA misuse?
Some of it. Scanners are good at structural checks: aria-hidden containing focusable elements, ARIA attributes that aren't allowed on a role, required child roles that are missing, invalid attribute values. Those are rule-based checks with clear answers, and the ACT rule above is written so tools can apply it the same way.
What a scanner can't judge is whether the ARIA is right. role="menu" on a set of site links passes every structural check if the required children are present. A tool can't tell you that a menu pattern was the wrong choice, or that an aria-label says something different from what the button does. That gap is the general one: across more than 2,000 audits, Deque found automated testing covered about 57% of issues by volume, against the common 20 to 30% estimate by success criteria (Deque, 2021). The rest needs a person. WebAIM says the same about its own numbers: "Absence of detected errors does not indicate that a page is accessible or conformant" (WebAIM Million 2026).
Our scanner runs axe-core, and it'll flag the structural ARIA failures with the selector and the WCAG criterion attached. It won't tell you your tab pattern is a bad idea. You need someone who uses a screen reader, or at least someone testing with one, for that.
Does an overlay that adds ARIA fix any of this?
I don't think so, and this is where I'm least neutral. An overlay that injects aria-label and role attributes at runtime is adding more ARIA, generated by a script that can't see what the component is meant to do. Every point above applies to it: the roles come without the behaviour, the labels can contradict the visible text, and the developer never sees any of it in the source. I'd rather fix ten elements in the markup than have a hundred patched invisibly. I go through what these scripts actually change on a page in what an overlay does to your page.
The legal frameworks look at the result, not the method. The EAA presumes conformity through EN 301 549, whose web chapter applies WCAG 2.1 AA, and WCAG 4.1.2 is in it. In the US, ADA web cases are argued with WCAG as the reference point. None of them give credit for attribute count.
Frequently asked questions
Does using ARIA make a website less accessible?
Not by itself, but misused ARIA does. WebAIM's 2026 data shows home pages with ARIA averaging 59.1 detected errors against 42 without, and WebAIM notes those pages are also more complex, so ARIA isn't proven to be the cause (WebAIM).
What is the first rule of ARIA?
Use a native HTML element when one has the semantics and behaviour you need, instead of adding ARIA to a generic element (W3C, Using ARIA). The Using ARIA note is now a discontinued draft kept for reference, and the formal author rules live in the ARIA in HTML Recommendation (W3C).
Is adding role="button" to a button harmful?
It's unnecessary rather than harmful in most cases. ARIA in HTML marks redundant roles as "NOT RECOMMENDED" because they can "potentially lead to unintended consequences" (W3C). Remove them.
Why is aria-hidden on a focusable element a problem?
Because a keyboard user can still reach it, but their screen reader announces nothing. The W3C says not to use aria-hidden="true" on focusable elements, and an ACT rule fails it under WCAG 4.1.2 (W3C ACT).
Can automated tools find all ARIA errors?
No. They find structural errors such as invalid attributes and hidden focusable content, but they can't judge whether a role or label is the right one. That needs manual testing with a keyboard and a screen reader.
See where your ARIA breaks
Run a free scan to find the structural ARIA failures on your pages, each with its selector and WCAG criterion. Then use the WCAG checklist to work through the criteria a scanner can't settle, and test your custom widgets with a keyboard and a screen reader before you call them done.
Written by Pavel Charkasau, founder of wcagc.com. Last updated 21 November 2026.
Sources
- WebAIM, The WebAIM Million: the 2026 report on the accessibility of the top 1,000,000 home pages (results from February 2026, last updated 30 March 2026): https://webaim.org/projects/million/ — accessed 21 November 2026.
- W3C WAI, ARIA Authoring Practices Guide, Read Me First: https://www.w3.org/WAI/ARIA/apg/practices/read-me-first/ — accessed 21 November 2026.
- W3C, ARIA in HTML, W3C Recommendation: https://www.w3.org/TR/html-aria/ — accessed 21 November 2026.
- W3C, Using ARIA (Discontinued Draft, 24 February 2026; the four rules of ARIA kept for historical purposes): https://www.w3.org/TR/using-aria/ — accessed 21 November 2026.
- W3C WAI, ACT rule 6cfa84, "Element with aria-hidden has no content in sequential focus navigation": https://www.w3.org/WAI/standards-guidelines/act/rules/6cfa84/ — accessed 21 November 2026.
- WHATWG, HTML Living Standard, the inert attribute: https://html.spec.whatwg.org/multipage/interaction.html#the-inert-attribute — accessed 21 November 2026.
- W3C, Web Content Accessibility Guidelines (WCAG) 2.2: https://www.w3.org/TR/WCAG22/ — accessed 21 November 2026.
- Deque, "Automated Testing Study Identifies 57% of Digital Accessibility Issues" (10 March 2021): https://www.deque.com/blog/automated-testing-study-identifies-57-percent-of-digital-accessibility-issues/ — accessed 21 November 2026.