Illustration: a magnifying glass inspects a button with keyboard focus in a browser window, next to an audit report with a score and three priorities, a Tab key, an NVDA label and an AA seal in front of the SiteCockpit logo on a light blue circle.

Accessibility Audit: Scope, Sampling and Report

What an audit tests, which pages belong in the sample and how to turn the report into measures for your online shop.

November 12, 2026 • 14 mins read

Summary

An accessibility audit combines automated analysis with manual expert testing and evaluates a representative sample against WCAG at level AA. According to the UK Accessibility Manual, tools alone find only about 30 to 40 percent of known barriers. Market surveillance under Annex 1 BFSG also works with samples, at least 10 percent of them chosen at random. A useful report names the location, severity and fix of each finding. As of 11/02/2026.

An automated scan delivers a score in minutes, while an audit delivers a reliable picture of where you actually stand. Anyone who wants to secure the accessibility of an online shop for the long term needs both, as our overview on checking for barriers correctly explains.

This article explains what an audit delivers, how the sample is chosen, which areas are tested and how to turn the report into an action plan. Research was completed on November 2, 2026.

What is an accessibility audit and what does it test?

An audit is a structured review that combines automated analysis with manual expert testing and measures the result against a fixed benchmark, usually the WCAG guidelines at level AA, increasingly in version 2.2. It does not ask whether a tool finds errors, but whether people using a keyboard, a screen reader or magnification can actually use your shop.

To draw the line: a scan is a partial automated check that runs in minutes and can be repeated as often as you like. An audit adds hands-on testing of a sample with keyboard, screen reader and magnification. A user test involves people with disabilities directly and shows whether a task succeeds in everyday use. All three serve the same goal but answer different questions, and none of them replaces the others.

The W3C provides the method with WCAG-EM. It divides the evaluation into five steps: define the scope, explore the website, select a representative sample, audit the sample and report the findings (W3C WAI). A new version, WCAG-EM 2, has been available since July 23, 2026 (W3C WAI).

The benchmark is changing as well. The European standard EN 301 549 V4.1.1 was published on September 2, 2026 and raises the web requirements to WCAG 2.2. Whether it has legal effect depends on its citation in the Official Journal, which according to specialist sources was still pending on October 17, 2026. We classified this in our article on EN 301 549 V4.1.1 for online shops. For planning purposes, a note from the European Commission helps: there are no breaking changes compared with version 2.1, so anyone testing against 2.2 largely covers the previous requirements as well.

Why is an automated scan not enough?

Because tools can only reliably test part of the criteria. The UK Government Digital Service compared tools on a page with 142 known barriers, and even the best ones found only about 30 to 40 percent (DWP Accessibility Manual). The US team CivicActions assigned 230 violations from its projects to testing methods: 12 could be confirmed by automation alone, 72 only manually and 146 needed both. A range of 30 to 60 percent is often cited, and the exact figure depends on how issues are counted and on the test page.

A practical example shows the gap. In the success story on mientus, the automated analysis in June 2024 initially returned 76 percent (level A), 75 percent (AA) and 73 percent (AAA). The manual expert review showed that actual conformance was only around 55 to 60 percent. Invisible to tools were, among other things, brand sliders and menus without keyboard operation, missing focus indicators, missing skip links and faulty ARIA attributes. This is a single project and not a statistic, but it illustrates where scans reach their limits.

This leads to a division of labor. The scan, for example with easyMonitoring, is the quick entry point and the continuous check. The audit settles the questions of judgment: Is an alternative text meaningful? Is the focus order logical? Can the checkout be completed with a screen reader? Only this kind of review says something about the usability of your shop for everyone.

Which rules govern how the sample is chosen?

A review cannot cover every page, so it works with a sample, and market surveillance follows the same principle. Annex 1 of Germany's Accessibility Strengthening Act (BFSG) sets out what the authority reviews on websites: home page, login, sitemap, contact, help pages and pages with legal information, plus the pages with the accessibility information required under Section 14 BFSG and selected example pages with a clearly different appearance or other types of content. Randomly chosen pages are added, amounting to at least 10 percent of this sample. If a page contains a step in a process, all steps of that process are checked. Under Section 28 BFSG, the authority may also review services without a specific reason, based on appropriate samples.

The W3C likewise describes structured and randomly selected pages for cases where not every page can be examined (W3C WAI). Our assessment: the structured selection makes sure every page type is included, while the random selection protects against reviewing only the well-maintained showcase pages. Only both together say something about the whole offering.

How the authority approaches this in practice is described in our article on market surveillance after more than a year. The logic carries over to an audit. The following selection for an online shop is our own derivation from Annex 1, not a legal requirement:

  • Home page and navigation: the central entry point with menus, search and skip links.
  • Category page with filters: product lists and filters are often heavily script-driven.
  • Product detail page: images, variants and the add-to-cart function.
  • Complete checkout process: every step from the cart to the confirmation, because accessibility is decided across the whole flow and not on the first page.
  • Login and customer account: forms with error messages.
  • Contact, help and legal pages: pages the authorities list explicitly.
  • The page with the accessibility information: it also belongs in the official sample.
  • Pages with a different look: for example promotion pages with dynamic prices.

The success story on mientus names a similar selection: home page, legal notice, login and registration, category page, product detail page and sale overview, each tested on desktop and on mobile.

Which test areas belong in an audit?

At least six, because each area places different requirements on the interface and covers a different kind of barrier. For contrast, the W3C names two thresholds: text needs at least 4.5:1, large text at least 3:1, and controls and important graphics need 3:1 against their surroundings. The values are thresholds and are not rounded, so 4.499:1 does not pass. Zoom and reflow require that text can be enlarged to 200 percent and that content stays readable at 400 percent zoom without scrolling in two directions (Accessibility Toolkit of the New South Wales Government).

Overview of the test areas of an audit
AreaWhat is testedWhy a tool is not enough
Keyboard and focusAll functions reachable without a mouse, focus visible, skip links presentWhether sliders, menus and dialogs are operable only shows in a hands-on walkthrough
Screen readerTest with NVDA and VoiceOver, names, roles and status messagesWhether content is read out in a meaningful way is a question of judgment
ContrastText at least 4.5:1, large text and controls at least 3:1Backgrounds with images and every state need a visual check
Zoom and reflow200 percent text enlargement, 400 percent zoom without two-dimensional scrollingLayout breaks only show up with real magnification
Forms and errorsLabels, understandable error messages, completion of the orderThe whole flow counts, not a single page
Structure and markupHeading hierarchy, semantic elements, ARIAWhether the structure fits the content is decided by a person

A scenario shows why hands-on testing remains necessary: the filter bar of a category page works with a mouse and passes every automated test. With a keyboard, however, tabbing skips past the filters, and a screen reader announces only buttons without names. For users without a mouse, the whole product list is then not usable, even though the requirements for keyboard operation apply to every function. Only a walkthrough with the assistive tools themselves finds cases like this.

The manual audit in the success story on DeinSportsfreund covered seven areas, including keyboard accessibility, screen reader compatibility, color contrast and zoom behavior. The W3C also recommends involving real users with disabilities in the evaluation, because they reflect the actual experience.

What does the report contain, and how does it become an action plan?

A useful report names the location, the affected criterion, the severity and a hint for the fix for every finding. The W3C provides a template that suggests which details an evaluation report should contain (W3C WAI). The SiteCockpit success stories show how this looks in practice: at DeinSportsfreund, the audit delivered a prioritized roadmap with concrete recommendations and code examples, while at mientus the barriers were sorted into three priority levels, with the most critical ones affecting basic operability.

Context matters here. According to the W3C, the methodology alone is in most cases not enough to declare conformance, because usually only a sample is evaluated. An audit is therefore a snapshot and not a guarantee. Typical findings show how concrete a report becomes. In the success story on mientus, forms offered no sufficient error validation, and the order process had no final review page. At DeinSportsfreund, more than 80 product images were missing descriptive alternative texts. The first is a task for development, the second for the editorial team, and both can be worked through in a targeted way. For implementation, we recommend this order:

  • Blocking errors first: everything that keeps keyboard or screen reader users from buying, above all in the checkout.
  • Template errors before individual errors: one fix in the theme affects many pages at once.
  • Turn editorial errors into a process: alternative texts and headings are created continuously and need fixed workflows.

You will find more on weighting in the article The costliest barrier mistakes and where companies should start. The results also feed into your accessibility information: Annex 3 BFSG requires a description of how the service meets the requirements. How to set up this information is explained on our page on the accessibility statement.

In-house or external: who should carry out the audit?

Both are possible. The W3C describes WCAG-EM explicitly for self-assessment and for evaluations by third parties, and calls the methodology independent of specific tools, browsers or assistive technologies (W3C WAI). An in-house audit makes sense if the team has experience with digital accessibility, screen readers and keyboard testing, and someone documents the approach.

An external review offers an independent view, experience with typical patterns such as sliders, filters and checkout, and a report that development and editorial teams can take over without extra effort. In both cases it is crucial that the benchmark and requirements are fixed before testing, so that later results remain comparable.

How often should an audit take place?

We recommend at least once a year and after major changes to the theme, checkout or apps. In between, ongoing monitoring should catch regressions, because a single review cannot do that. This is a recommendation and not legal text: Section 14 (3) BFSG merely requires that the service provider always meets the requirements. An audit shows the state at one point in time, and any change can make accessibility worse again.

On effort: according to the product page on manual audits, an audit takes between a few days and a maximum of two weeks, depending on the size of the website. The existing site does not need to be prepared for it.

SiteCockpit: from audit to lasting oversight

An audit is not a one-off project but the starting point for lasting accessibility. SiteCockpit combines several building blocks for this:

  • Manual audits: experts add a manual assessment to the results from easyMonitoring and deliver a report with recommendations.
  • easyMonitoring: checks your pages regularly, prioritizes the need for action and makes progress measurable in the dashboard.
  • easyStatement: creates the accessibility statement based on real test results.
  • Livecheck: a free first assessment in under 60 seconds.

The success stories on DeinSportsfreund (first scan, manual audit, ongoing optimization) and on mientus show how this works together.

What makes sense now

An audit replaces neither ongoing monitoring nor editorial upkeep, but it provides the most honest status. If you first need only an overview, start with a scan and plan the audit for your most important pages.

  • Establish the status: clarify who is affected by the law and which requirements apply to your shop, define the sample according to Annex 1 and get a first overview with the Livecheck.
  • Test in depth: carry out the audit for the sample, in-house following WCAG-EM or with external support.
  • Keep it in view: prioritize the report, plan measures and set up monitoring.

A step-by-step approach is enough: first the checkout, then the remaining page types. With easyMonitoring you keep track of progress along the way.

How accessible is your website?

Test it for free with the SiteCockpit Livecheck and get a first assessment in under 60 seconds.

Test now for free

Table of Contents

View blog articles
Sources
(17) SiteCockpit: Success story mientus Retrieved on 11/02/2026
(19) SiteCockpit: Manual audits Retrieved on 11/02/2026