Iron Goo
Iron Goo robot tests a keyboard and color swatch beside a monitor, with an unplugged mouse on the desk

Website Accessibility Checklist for Every Site

Atamyrat Hangeldiyev
Atamyrat Hangeldiyev
Systems Architect
March 9, 2026
On this page
UX

The power of the Web is in its universality. Access by everyone regardless of disability is an essential aspect.

W3C WAI, Introduction to Web Accessibility, quoting Tim Berners-Lee

Our website accessibility checklist is a short set of MUST, SHOULD and AVOID requirements tied to WCAG 2.2 Level A and AA, each with the customer it can fail and a test you can run yourself. The most useful starting fact: the WebAIM Million examination of one million home pages in February 2026 detected WCAG 2 failures on 95.9% of them, averaging 56.1 detected errors per page. Six categories accounted for 96% of detected errors.

Our opinion: accessibility is the same work as good UX. Each requirement represents someone who may struggle to finish a booking, ask for a quote or pay. The wider customer journey and UX includes the steps before and after those tasks too.

WCAG 2.2 became a W3C Recommendation on 5 October 2023; the current published Recommendation is dated 12 December 2024. It added nine success criteria and removed 4.1.1 Parsing as obsolete. The checks here are a practical starting set, not every criterion needed for conformance.

MUST means a failure can block a task outright on a page where customers book, request a quote or pay. SHOULD means a failure can make that task slow or error-prone. This is our priority system: every applicable Level A and AA criterion remains part of Level AA conformance, regardless of its place on this list.

95.9%
Home pages with detected failures
4.5:1
Minimum contrast for ordinary text
24 by 24
Minimum target size in CSS pixels
29%
Barriers missed in an older tool study
Form with hidden focus, a small target, a red-only error and an unnamed icon button

Notice how four separate defects can interrupt the same form: hidden focus, a small target, an error communicated only by color and an icon without an accessible name.

MUST: make the main task operable and understandable

A1. 1.1.1 Non-text Content (Level A): give useful alternatives

Requirement. Give meaningful images text alternatives that serve the same purpose. Give icon-only controls an accessible name describing their action; decorative images can be skipped by assistive technology.

Why. A customer who cannot see an image needs its relevant information in text. Someone using a screen reader cannot choose an unnamed magnifying-glass button with confidence.

Test. Turn on VoiceOver or NVDA and move through images and icon controls. Pass if meaningful images have useful alternatives, decorative images are skipped, and every icon control announces its purpose.

A2. 1.4.3 Contrast (Minimum) (Level AA): make text readable

Requirement. Ordinary text needs at least 4.5:1 contrast against its background. Large-scale text needs 3:1: at least 18 point, or 14 point bold. Logotypes and incidental or inactive text are exempt.

Why. A customer with low vision may be unable to read pale form instructions. Bright sunlight can also make a weak label difficult to see on a phone.

Test. In Chrome DevTools, inspect the text and open the color picker to read its contrast ratio. Pass at 4.5:1 for ordinary text or 3:1 for qualifying large-scale text.

A3. 1.4.11 Non-text Contrast (Level AA): show controls and states

Requirement. Visual information needed to identify a component or its state needs at least 3:1 contrast against adjacent colors. Check input boundaries and focus indicators as well as the visible parts of controls.

Why. A customer with low vision may not see where an empty field starts or which control has focus. A label does little good if the field itself disappears into the background.

Test. Use the Chrome DevTools color picker to compare each necessary boundary or state with its adjacent color. Pass when the ratio is at least 3:1, accounting for the criterion's exceptions.

A4. 2.1.1 Keyboard and 2.1.2 No Keyboard Trap (Level A): complete and leave every control

Requirement. Make functionality operable through a keyboard without requiring timed keystrokes, subject to the path-dependent input exception. If focus enters a component by keyboard, the user must be able to leave it with the keyboard.

Why. A customer who cannot use a mouse may reach a booking form and then get stuck inside a menu. That is a stopped task, even if the page looks polished.

Test. Use Tab, Shift+Tab, Enter, Space, arrow keys and Escape through the main task. Pass if you can reach and operate every necessary control and leave each widget without a mouse.

A5. 2.4.7 Focus Visible (Level AA): show where the keyboard is

Requirement. Provide a visible keyboard focus indicator. Any keyboard-operable interface needs a mode of operation in which the keyboard focus indicator is visible.

Why. A keyboard user needs to know which button will activate before pressing Enter. Without a visible indicator, a form becomes a guessing exercise.

Test. Press Tab and Shift+Tab through the page. Pass if each focused control has a visible indicator; fail if focus moves and you cannot see its location.

A6. 2.4.11 Focus Not Obscured (Minimum) (Level AA): keep focused controls in view

Requirement. A focused component must not be entirely hidden by author-created content. Sticky headers, cookie banners and chat widgets are common places to check.

Why. A keyboard user may tab to a submit button that sits behind a fixed banner. The control can technically have focus while the customer cannot see where it is.

Test. Tab through the page at a normal desktop and phone-sized viewport with banners and chat controls present. Pass if no focused component is entirely hidden by those elements.

A7. 3.3.1 Error Identification and 3.3.2 Labels or Instructions (Level A): identify fields and errors

Requirement. Give inputs labels or instructions. When an input error is detected, identify the affected item and describe the error in text.

Why. A customer using a screen reader needs to know which field failed; a customer who cannot distinguish the red border needs the message too. If several fields regularly cause trouble, revisit the form's conversion path.

Test. Submit the form with a required field empty and a phone number in the wrong format. Pass if labels identify the inputs and text identifies each detected error and its field.

A8. 4.1.2 Name, Role, Value (Level A): expose controls to assistive technology

Requirement. Make each user interface component's name and role programmatically determinable. User-settable states, properties and values must be programmatically settable, and changes must be available to user agents, including assistive technology. Standard HTML controls meet this criterion when used according to specification.

Why. A customer using a screen reader needs to hear that a control is a button and what it does. A custom control announced only as “button” leaves the action unknown.

Test. With VoiceOver on Mac or free NVDA on Windows, tab through controls and change their states. Pass if each announces an appropriate name and role, and relevant state changes are announced.

SHOULD: reduce errors and unnecessary effort

A9. 2.5.8 Target Size (Minimum) (Level AA): leave room to tap

Requirement. Pointer targets should be at least 24 by 24 CSS pixels. For the spacing exception, a 24 CSS pixel diameter circle centered on each undersized target must not intersect another target or another such circle. The criterion also has Equivalent, Inline, User agent control and Essential exceptions.

Why. A customer with a tremor may hit the neighboring control when targets are small and crowded. The same check supports the mobile side of the customer journey.

Test. Inspect the target in Chrome DevTools and read its box-model width and height in CSS pixels. Pass at 24 by 24 or larger, or verify that a named exception applies.

A10. 3.3.3 Error Suggestion (Level AA): tell people how to correct an error

Requirement. When an input error is detected and a correction is known, provide a suggestion unless it would jeopardize security or the purpose of the content. The wording should identify a usable next action.

Why. A customer who enters a phone number in an unsupported format may otherwise try several guesses. Good error messages and label text make the correction explicit.

Test. Enter a wrong-format phone number and submit. Pass if the detected error includes a useful format suggestion when one is known and no exception applies.

A11. 3.3.7 Redundant Entry (Level A): carry information through a process

Requirement. Information already entered or provided in the same process, and required again, must be auto-populated or available to select. Exceptions apply when re-entry is essential, needed for security or the earlier information is no longer valid.

Why. A customer with a cognitive or motor disability may find repeated entry tiring or error-prone during a multi-step checkout. This criterion is Level A, though it is sometimes incorrectly listed as AA; it sits under SHOULD here because it is most relevant to multi-step forms.

Test. Complete a multi-step process that asks for the same information twice. Pass if the later step fills it or offers it for selection, unless a stated exception applies.

A12. 3.3.8 Accessible Authentication (Minimum) (Level AA): provide a workable way to sign in

Requirement. Do not require a cognitive function test, such as remembering a password or solving a puzzle, at an authentication step unless an allowed alternative or assisting mechanism is available. Object recognition and personal content are also permitted alternatives. W3C names password manager support and copy and paste as assisting mechanisms.

Why. A customer who cannot remember a password or solve a puzzle CAPTCHA needs another way through authentication. A sign-in wall can otherwise prevent an accessible checkout from being reached.

Test. Paste into password and one-time-code fields, try a password manager, and inspect any authentication CAPTCHA for an allowed alternative or mechanism. Pass if the authentication flow supplies an allowed route without requiring the customer to solve an unsupported cognitive test.

AVOID: choices that undermine the checks

A13. 3.3.8 Accessible Authentication (Minimum) (Level AA): blocking paste

Requirement. Our reading: avoid blocking paste in password and one-time-code fields. Copy and paste can assist a person with an authentication task.

Why. A customer who stores credentials in a password manager or retrieves a code from another application may struggle to transcribe them. Blocking paste removes a useful mechanism.

Test. Copy a password and a one-time code, then paste each into its field. Pass if both fields accept paste; also assess the full authentication flow against A12.

A14. 2.4.7 Focus Visible (Level AA): removing the focus outline

Requirement. Avoid removing the focus outline unless another visible focus indicator takes its place. A custom indicator must still let keyboard users locate focus.

Why. A customer using only a keyboard can lose their place when the browser outline disappears. They may activate the wrong control or abandon the task.

Test. Tab through links, fields and buttons. Pass if focus remains visible on every control, including those with custom styling.

A15. 3.3.1 Error Identification (Level A): showing an error only as a red border

Requirement. Avoid using a red border as the only indication of an input error. Our reading of the criterion is that the detected error needs a text description identifying the affected item.

Why. A customer who cannot distinguish the border color, or who uses a screen reader, may receive no explanation of the failed submission. The field and the problem both need to be identifiable.

Test. Submit a required field empty. Pass if text identifies the field and describes its error; a red border alone fails.

A16. 4.1.2 Name, Role, Value (Level A): relying on an overlay

Requirement. Avoid treating an overlay widget as the fix for inaccessible underlying controls. In our opinion, repairing the components is the dependable route because the markup and behavior the criteria test need to work.

Why. A customer using assistive technology still needs a control with a usable name and role. A widget's presence does not by itself establish that the underlying control provides them.

Test. With the overlay present, use VoiceOver or NVDA to reach and operate the underlying controls. Pass this spot check only if their names, roles and relevant states are available; a widget appearing on screen is not a pass.

A17. Treating an automated pass as conformance

Requirement. Avoid declaring conformance from a clean automated report. Test whether alternatives actually communicate the purpose of the content.

Why. A customer can encounter an image with an unhelpful text alternative even when a tool detects an alt attribute. The attribute's presence does not establish that its wording works.

Test. Run axe DevTools, Lighthouse or WAVE, then manually compare meaningful images with their alternatives. Pass this check only if the alternatives serve the images' purposes; a clean report alone is insufficient.

Requirement reference

IDCriterionLevelTest
A11.1.1 Non-text ContentAHear useful alternatives and names.
A21.4.3 Contrast (Minimum)AAMeasure text contrast in DevTools.
A31.4.11 Non-text ContrastAAMeasure required boundaries and states.
A42.1.1 Keyboard; 2.1.2 No Keyboard TrapAOperate and leave controls by keyboard.
A52.4.7 Focus VisibleAASee focus at every tab stop.
A62.4.11 Focus Not Obscured (Minimum)AACheck focused controls against fixed content.
A73.3.1 Error Identification; 3.3.2 Labels or InstructionsASubmit empty and wrong-format fields.
A84.1.2 Name, Role, ValueAHear names, roles and state changes.
A92.5.8 Target Size (Minimum)AAMeasure targets and check exceptions.
A103.3.3 Error SuggestionAASubmit a wrong-format value.
A113.3.7 Redundant EntryACheck repeated information in one process.
A123.3.8 Accessible Authentication (Minimum)AATest paste, password manager and CAPTCHA.
A133.3.8 Accessible Authentication (Minimum)AAPaste a password and one-time code.
A142.4.7 Focus VisibleAACheck focus after custom styling.
A153.3.1 Error IdentificationACheck that an error is described in text.
A164.1.2 Name, Role, ValueAInspect underlying controls with a screen reader.
A17None (testing practice; see 1.1.1 for alt quality)n/aReview alternatives after an automated run.

What automated accessibility tools miss

WebAIM's six common categories are a useful place to start because tools can detect many instances. In its February 2026 examination, low contrast appeared on 83.9% of home pages, missing image alternatives on 53.1%, missing form labels on 51%, empty links on 46.3%, empty buttons on 30.6% and missing document language on 13.5%. These percentages describe home pages with each detected issue; they are not shares of all errors. Accessibility is one part of the whole customer journey and its user experience.

The older Government Digital Service study from 24 February 2017 tested ten tools against 143 deliberate barriers. All tools missed 42 barriers, or 29%; together they found 71%. The best single tool found 37% through errors and warnings, the best found 41% when manual inspection prompts were included, and Google Developer Tools found 17%.

Those results describe the tools and test page used in 2017, not a measurement of today's versions. Separately, the axe-core README claims that its tool finds an average of 57% of WCAG issues automatically; that is Deque's own claim.

A clean report is not conformance

Automated detection has limits. An absence of detected errors does not establish accessibility or WCAG conformance, so inspect the task with a keyboard and assistive technology as well.

A 30-minute accessibility test session

Choose a page with a customer task and record each failure with its URL, control and steps to reproduce it. Thirty minutes is a spot check, not a full conformance evaluation.

Two people test the same page using a keyboard and a screen reader

One tester follows the task using the keyboard while another spot checks the same page with a screen reader. Compare what each person can reach and identify.

  1. →
    First 10 minutes: keyboard

    Set the mouse aside and use Tab, Shift+Tab, Enter, Space, arrows and Escape to attempt the main task. Record any unreachable control, invisible or fully obscured focus, and any widget you cannot leave.

  2. →
    Next 5 minutes: contrast and size

    In Chrome DevTools, use the color picker to check text and required control contrast. Inspect target dimensions in the box model and record ratios below the applicable threshold or undersized targets needing an exception check.

  3. →
    Next 5 minutes: automated run

    Run axe DevTools, Lighthouse or WAVE on the page. Save the findings with the page URL and check each reported issue; an empty report is only the start of manual review.

  4. →
    Last 10 minutes: screen reader

    On a Mac, start VoiceOver with Cmd+F5; on Windows, use free NVDA. Tab through the controls and listen for a meaningful name and role, then check whether relevant state changes are announced.

If the same custom controls and missing labels appear across the template, a component-level repair can address the repeated source. Foundation is a site rebuild on a current stack; its two-week audit phase scores the current site's performance, accessibility and structured data, with a target of a 90 plus mobile Lighthouse score.

This is general information, not legal advice. Sources were read on 29 October 2026.

The European Accessibility Act, Directive (EU) 2019/882, applies from 28 June 2025 to listed products and consumer services, including e-commerce and consumer banking. Article 4(5) exempts microenterprises providing services from the service accessibility requirements and related obligations.

Article 3(23) defines a microenterprise as employing fewer than 10 people and having annual turnover or an annual balance sheet total not exceeding EUR 2 million. A company with 10 or more people therefore is not a microenterprise under that definition. Article 32 has Member States allow a transitional period to 28 June 2030 during which service providers may keep using products they lawfully used to provide similar services before 28 June 2025; qualifying service contracts agreed before 28 June 2025 may continue unaltered until expiry, for no longer than five years.

In the US, the DOJ Title II web rule, published on 24 April 2024, sets WCAG 2.1 Level AA for state and local governments. The 20 April 2026 interim final rule moved compliance dates to 26 April 2027 for governments with a total population of 50,000 or more and 26 April 2028 for those with a total population of less than 50,000, or any special district government. Government content must meet the rule even when a private firm builds or runs it.

For businesses open to the public under Title III, DOJ has not adopted a technical standard for private business websites. Its 2022 web guidance says businesses have flexibility in meeting ADA requirements and names WCAG as helpful guidance.

Start today by opening your booking, quote or payment page and completing its main task with the keyboard. Write down the first control you cannot reach, operate or see, then give that failure to the person responsible for the site.

Free ebook

Put your next steps in order

Build on what you have just read with a free ebook on semantic SEO and AI visibility. Get worked examples, decision rules, checklists and a do-this-week action in every chapter, plus a six-week plan for your site.

By signing up and confirming your address, you consent to receive the free ebook plus one email every two weeks with 3 or 4 new Iron Goo videos, guides and articles, and nothing else; unsubscribe any time with one click.

Related in UX

Ready to move?

Tell us where your business is today. A written assessment lands in your inbox within two business days.

Talk to us