Test a website for accessibility

This evaluation can say something about 27 of the 50 success criteria — to differing depth. The pages are loaded in a real browser for this; what is checked is the rendered page, not the source code: The list also includes 2 criteria from WCAG 2.2, which are not counted.

• 1.1.1 Non-text Content — checked broadly

  Whether a text alternative exists can be decided by machine. Whether it describes the image accurately cannot.
• 1.2.2 Captions (Prerecorded) — checked in one specific respect

  A video element without a caption track is detected. Embedded players and burned-in captions remain invisible.
• 1.3.1 Info and Relationships — checked broadly

  Labels, lists and tables are checked. Whether a visual structure was marked up at all is not measurable.
• 1.3.5 Identify Input Purpose — checked in one specific respect

  The validity of autocomplete values that are present is checked. A missing value on a personal-data field cannot be decided.
• 1.4.1 Use of Color — checked in one specific respect

  Links within body text are checked. Colour as the sole carrier of meaning elsewhere remains open.

• 1.4.2 Audio Control — checked completely

• 1.4.3 Contrast (Minimum) — checked broadly

  Over gradients, background images and semi-transparent surfaces the contrast cannot be computed reliably; those places appear as "to review".
• 1.4.4 Resize Text — checked broadly

  Page zoom to 200% and a zoom lock in the viewport declaration are measured. Text-only resizing without a layout change is not checked.

• 1.4.10 Reflow — checked completely

• 1.4.12 Text Spacing — checked completely

• 2.1.1 Keyboard — checked in one specific respect

  Controls that cannot receive focus are detected. Whether a reachable function is also operable only a keyboard walk-through shows.
• 2.1.2 No Keyboard Trap — checked broadly

  The case where focus cannot be released at all is found. Whether a deliberately bound focus — a modal dialog, a consent banner — is bound correctly is presented for review.
• 2.2.1 Timing Adjustable — checked in one specific respect

  Timing via meta refresh is detected. JavaScript timers and session timeouts remain open.
• 2.2.2 Pause, Stop, Hide — checked in one specific respect

  Two obsolete elements are detected. Auto-updating content driven by CSS or JavaScript remains open.
• 2.4.1 Bypass Blocks — checked broadly

  Whether a bypass mechanism exists is checked - not whether it skips the right blocks.
• 2.4.2 Page Titled — checked broadly

  A title must exist - whether it describes the topic or purpose of the page is a judgement.
• 2.4.3 Focus Order — checked in one specific respect

  Positive tabindex values are detected. Whether the order is coherent is asked as a question.
• 2.4.4 Link Purpose (In Context) — checked broadly

  A link without an accessible name is detected. Whether an existing name identifies the target is a judgement.
• 2.4.5 Multiple Ways — checked in one specific respect

  A navigation region, search field, site map and breadcrumb trail are detected. Fewer than two found is presented for review as a manual check — never reported as an automatic violation. A search behind a toggle, or a route through an image map, is not detected.
• 2.4.7 Focus Visible — checked broadly

  What is measured is whether anything changes visibly on focus. Whether the change is prominent enough to be noticed is not decided by machine.
• 2.4.11 Focus Not Obscured (Minimum) — checked in one specific respect · additionally checked under WCAG 2.2

  Only complete obscuring is reported, only at the focus stops reached, and only in the desktop viewport.

• 2.5.8 Target Size (Minimum) — checked completely · additionally checked under WCAG 2.2

• 3.1.1 Language of Page — checked completely

• 3.1.2 Language of Parts — checked in one specific respect

  The validity of lang attributes that are present is checked. A missing lang on a foreign-language passage cannot be detected by machine.
• 3.2.1 On Focus — checked in one specific respect

  Page changes, new windows and submitted forms on focus are detected. Other changes of context cannot be recognised as such by machine.
• 3.2.3 Consistent Navigation — checked broadly

  The order of the entries inside repeated navigation regions is compared across the pages evaluated. Repeated controls outside such regions remain open.
• 3.2.4 Consistent Identification — checked in one specific respect

  Identical targets carrying different names within navigation regions are compared. Identical function without an identical target cannot be recognised by machine.
• 3.3.2 Labels or Instructions — checked in one specific respect

  A field with multiple labels is detected. Whether instructions are sufficient is a judgement.
• 4.1.2 Name, Role, Value — checked broadly

  Name, role and state are checked where they appear in the markup. Custom controls without any markup go unnoticed.

For every finding you see the affected place, a screenshot and the measured value — the actual contrast ratio, for instance. Whatever is listed above as “broadly” or “in one specific respect” stays open and goes into the statement as unchecked — not as passing.

23 of the 50 success criteria cannot be measured at all, and 23 more only in part. For every one of them the application guides you with a prepared question that you answer with “met”, “not met” or “not applicable” — grouped by what you have to do:

• Media and motion — 7 questions
• With the keyboard — 8 questions
• Forms and input — 6 questions
• Content and language — 8 questions
• Structure and assistive technology — 5 questions
• Presentation — 4 questions
• Pointer and gestures — 4 questions
• Across pages — 3 questions

A few screening questions save you whole blocks: if you have no videos, you are not asked about captions. Your answers feed into the statement and carry over to the next scan. Whatever you leave open is reported as not evaluated and never counted as passed. Your own assessment does not replace a review by third parties — the report marks it as such.

No. We read your pages like a visitor and write nothing back. There is no script to embed and no overlay — overlays do not remove barriers, often make operation with assistive technology worse, and are not legally recognised as evidence.

Not for the first scan. You only need an account when you want to answer the manual checks, check more pages, have checks run regularly or publish a statement — to publish, you have to prove that the domain is yours.

Without an account we check the same website at most 3 times a day. That is not a sales tactic but consideration: every check loads dozens of pages from someone else’s server, and we cannot know whether you run it. Once your domain is verified, this limit does not apply to your own website.

The results of a free scan, including screenshots, are deleted after 7 days. Form field contents are masked in the testing browser before anything is stored.

No. You receive test data and a text template, not a legal assessment of your particular case. A conclusive assessment still requires human review — but we show you exactly which points that concerns.