Accessibility

01 What we set out to do

malziME is a learning tool for media literacy, and it is aimed at schools. Anyone who teaches about data protection has to do a clean job on usability too. That is why we tested malziME against the international standard WCAG 2.2, Level AA – by measuring, not by asserting.

We are not obliged to. The Austrian Accessibility Act (Barrierefreiheitsgesetz, BaFG), applicable since 28 June 2025, exempts micro-enterprises providing services under § 6(1) – that is, businesses with fewer than 10 employees and no more than 2 million euros in annual turnover or balance sheet total. malziland is a one-person business and falls under that exemption.

We do it anyway, because our audience is schools – public bodies with an obligation of their own – and because a tool that excludes people misses its purpose.

02 Conformance status

03 What was tested – and how

Testing followed WCAG-EM 2.0, the official evaluation methodology of the W3C. It prescribes what an evaluation report must contain – among other things, which devices a page is expected to work with, and that a multi-step process is tested at every step, not only in places.

All pages were tested in all essential states – 70 states per browser: empty home page, image preparation, queue, AI output in progress, finished profile in serious and in Beast mode, reality check, PDF output, error messages, connection loss, opened tooltips.

• Three browser engines: Chromium (Chrome, Edge), WebKit (Safari) and Gecko (Firefox) – which covers every engine that can run on participants' devices.
• Two screen sizes: 1280 pixels on a computer and 320 pixels on a small phone.
• Tool: axe-core, restricted to the binding WCAG rules – no optional recommendations that would flatter the result.
• Our own measurements for target sizes, reflow on narrow screens, text scaling to 200 %, altered text spacing and the visibility of the keyboard focus.
• Every measurement runs twice. Only what occurs both times is recorded – otherwise measurement errors end up in the report as findings.
• Where the tool abstains, we measure again. axe cannot judge some colour contrasts and reports them as „cannot tell“. That is not a pass. In such cases we photograph the element and recalculate the contrast pixel by pixel. That is exactly how we found and fixed a too-faint info symbol on 17 August.
• Every measurement has a control. If the testing tool finds nothing to test, the run aborts. Otherwise a broken run would look like a perfect result. On 19 August that is exactly what happened: a popup on the map covered the profile content in such a way that the tool never looked at it – and reported “no violations” without having looked. Behind it were three text colours in Beast Mode that were too pale; all three were fixed the same day. Since then the check first states what it looked at, and raises an alarm when a visible area is missing.

The full evaluation report covering all 55 criteria, the method used and the result for each criterion is available on request.

04 Known limitations

We name them rather than keeping quiet about them.

• The example photos contain text as part of the image. The label „KI erstellt“ or „AI generated“ is burned into the pixels. Article 50 of the EU AI Act has applied since 2 August 2026 and requires artificially generated image content to be disclosed – expressly including images of people who do not exist. We put the label into the pixels because a label placed beside the image disappears as soon as someone saves it and passes it on. For people who cannot see, it is also in the alternative text.
• We do not aim for the highest level, AAA. The W3C expressly does not recommend it as a requirement for entire websites. We nevertheless meet ten AAA criteria; six are demonstrably not met – among them the enhanced contrast ratio of 7 : 1, which would require reworking our brand colours, and the reading level criterion, which calls for a second, simplified version of all legal texts.
• Not all assistive technologies have been tested. Why this leads to the statement „partially conformant“ is explained in section 02. If you work with assistive technology: your feedback is worth more to us than any further measurement.
• The embedded map comes from OpenStreetMap and is not designed by us. Its content is also available as text.

05 What we deliberately do not use

• No accessibility overlay. Tools that promise accessibility at the click of a button and supply a badge with it demonstrably make things worse for people using a screen reader. We regard such a badge as a warning sign, not as a credential.
• No purchased seal. We looked into external certification and decided against it. A traceable evaluation report with openly named gaps says more than a certificate without a method.

06 Found something? Please tell us.

If you run into a barrier, write to us:

We reply within two weeks. Please describe, if you can, which page is affected and what device or assistive technology you were using. Even a brief note helps – we will keep looking ourselves.

Should you feel disadvantaged by a barrier, an independent conciliation procedure is available to you at the Sozialministeriumservice (Austrian Social Ministry Service). It is free of charge and informal, is conducted by trained conciliation officers, and any interpreting costs are borne by the Sozialministeriumservice. Applications can be made online or on paper via sozialministeriumservice.gv.at.

For completeness: this procedure belongs to Austrian disability equality law and concerns discrimination. It is not the complaints procedure under the Accessibility Act – which does not apply to us as a micro-enterprise. We mention it because we do not want to hide the route, not because we are obliged to.

07 About this statement

Prepared on 17 August 2026, last reviewed on 19 August 2026, on the basis of a self-assessment using the tools listed under 03. We test again whenever the appearance, the interaction or the page structure changes, and at least every six months. The automated measurements run with every release.