ellaccess tests digital accessibility in days, not weeks

Accessibility is a legal obligation for public bodies and, since June 2025, for many companies. Yet manual audits take weeks and expert capacity is scarce. Our AI agents test websites, shops and specialist applications the way human auditors do: criterion by criterion, with annotated evidence and concrete fixes for editorial, product and IT teams.

Built in the context of German federal IT. It complements human testing expertise instead of replacing it.

Request a free initial assessment How it works

98

test steps make up the BIK web test, the established German procedure for auditing digital accessibility. Today it is worked through largely by hand

10–12

person-days per manual audit, while expert testing capacity stays scarce

7,508

websites and apps of public bodies were reviewed by the federal monitoring body (BFIT-Bund) from 2022 to 2024. Not a single one met all requirements in full

Source: bfit-bund.de

€100,000

is the maximum fine under the German Accessibility Act (BFSG) for certain violations. The market surveillance authority also checks without specific cause

Source: § 37 BFSG
Process

How ellaccess works

1

Define the scope

Domain, URLs and screens, from government portal to online shop.

2

Agents test like auditors

Playwright-based agent sessions work through forms, workflows and states, criterion by criterion along the BIK methodology.

3

Evidence per criterion

Annotated screenshots and findings make every result verifiable. No black-box score.

4

Actionable report

A report with concrete fix recommendations for editors and IT. Re-test any time: before go-live and after every release.

The result

A report your teams can act on

Each finding maps to its test step and location and includes a concrete fix, ready for editorial and IT. We validate the agent against real expert audit reports: it finds largely the same issues, and across repeated runs additional ones a single pass can miss.

Test report: beispiel-portal.de

Example
  • Fail

    9.1.3.1 Info and relationships

    Multi-step order form, step 2: labels not programmatically linked

    → Link labels to fields via for/id

  • Fail

    9.1.4.3 Contrast (minimum)

    Footer links fall below 4.5:1 against the background

    → Darken link color to reach ≥ 4.5:1

  • Fail

    9.2.4.4 Link purpose (in context)

    Repeated "read more" links without context

    → Make link texts specific

Every finding ships with an annotated screenshot as evidence.

Why scanners aren't enough

Rule-based tools flag where something might be wrong: syntactically, with many false positives. So far, not a single BITV test step could be assessed fully statically. The BIK methodology demands operation: forms, workflows and states, per criterion with evidence.

Rule-based scanners

provide syntactic hints for roughly 30–40 test steps; none can be assessed fully statically

○ hint only, needs human validation

Agentic testing

assesses the 60 website-relevant WCAG test steps semantically, operating pages, with evidence per criterion

● assessed semantically · grey: annex steps, rarely relevant for websites

Illustrative view. Basis: the 98 test steps of the BIK web test (60 WCAG for websites + 38 annex): a testing method, not the letter of the law. What's owed are the applicable requirements.

Who it's for

Relevant for BITV 2.0 and the BFSG

Public sector

BITV 2.0 · state law · EN 301 549

  • Cities, municipalities & districts

    Portals publish new content daily. Conformance decays continuously. Test continuously instead of every few years.

  • State & federal authorities

    Born in the context of federal IT, complex specialist applications included. Verifiable evidence for acceptance and operational handover.

  • Municipal IT service providers

    Dozens of client portals and specialist applications with one repeatable testing process instead of separate audits per client.

Private sector

BFSG · European Accessibility Act

  • E-commerce & online retail

    Ordering, booking, paying: the federal agency reads the BFSG as covering the whole connected website, not just the checkout, though this is not settled law. We test the full consumer path.

  • Banks & payment services

    Consumer banking services fall under the BFSG: online banking and authentication explicitly; application flows insofar as they serve account, payment, credit or securities services.

  • Transport & telecommunications

    For passenger transport, the BFSG covers air, long-distance coach, rail and waterborne services including e-tickets and real-time information, plus telecommunications. Event ticketing counts as e-commerce once consumers conclude contracts online.

  • Media & e-books

    E-book services and access to streaming platforms are explicitly covered by the BFSG, from website and apps to the payment flow.

Frequently asked questions

Does the agentic test replace an expert audit? +
There is no official certificate in either regime: public bodies must be conformant and publish an accessibility statement (§ 12b BGG), companies under the BFSG must be conformant and provide the required accessibility information. The agentic test covers the same methodology and is repeatable throughout development. For the final assessment you can still bring in a human testing body, with far fewer iterations.
Does the BFSG apply to our website? +
If consumers can order, book, initiate contracts or authenticate on your website or app, it usually qualifies as e-commerce under the BFSG, in force since 28 June 2025. Market surveillance (MLBF) also checks without specific cause; non-conformance can lead to corrective orders up to prohibition of the service. Microenterprises are exempt for services. We clarify your scope in the free initial assessment.
What makes this different from rule-based scanners? +
Scanners provide syntactic hints for part of the criteria, with many false positives and no semantic assessment. Our agents execute the human test methodology: they navigate pages, fill in forms, operate workflows and assess each criterion with evidence.
Can login-protected specialist applications be tested? +
Yes, with test credentials you provide. Agent sessions use them to authenticate and work through stateful workflows. That is where most audit effort accrues today.
Which legal frameworks does the test cover? +
BITV 2.0 and state regulations for public bodies, the BFSG for private businesses. The practical benchmark is EN 301 549 with WCAG 2.1 AA. What's owed are the requirements applicable to your offering; our methodology follows the test steps of the BIK web test.
Can the test run on German infrastructure? +
Yes. For procurement and privacy requirements the test can run on German or European infrastructure, depending on the model with some trade-offs in speed.
When is the right moment to test? +
Whenever evidence is due: acceptance, go-live, operational handover, a funding or procurement proof, a monitoring finding or a complaint, and after every major release. That is exactly what agents make repeatable, instead of turning testing into one expensive one-off.
What does a test cost? +
Pricing depends on the number of URLs and screens in scope. Request a free initial assessment. We respond within one business day.
Contact

Request a free initial assessment

Tell us which website, shop or application we should look at. We respond within one business day to set up a short scoping call.