An AI that summarily turns Peggy into „Sonja“ makes for an amusing moment at first. At the same time, this mistake clearly illustrates what this episode is all about: Artificial intelligence can be fast, powerful, and helpful—but it is not infallible.
Precisely for this reason, its deployment requires human control, professional expertise, and clear responsibility. In this episode of „Finally Accessible,“ Peggy Reuter-Heinrich discusses the EU AI Act and the human-in-the-loop principle. At the heart of the discussion is the question of what consequences the increasing use of AI has for public authorities, municipalities, public institutions, and anyone procuring digital products or services.
Peggy takes a particularly critical look at supposed one-click solutions for digital accessibility:
AI overlays,
automated scanners,
and fully automatic PDF converters promise quick compliance.
Yet a green audit result is a far cry from ensuring that an offering is actually usable for people with disabilities. Using the example of complex PDF tables, it becomes clear how automated systems can seemingly eliminate formal errors while destroying structure, reading order, and comprehensibility.
The core message of this episode:
AI is exceptionally well-suited for preliminary checks, routine tasks, and identifying potential anomalies. However, the final evaluation, validation, and prioritization must be carried out by qualified humans.
It is precisely this interplay that HeiReS also deploys in the 2026 Smart City Study: AI assists with the preliminary check across numerous WCAG and inclusion criteria, after which experienced experts review the results and develop concrete recommendations for action.
In this episode, you will learn:
why AI results should never be accepted without review,
why automated accessibility promises must be critically questioned,
how testing programs can be seemingly satisfied by technically altered structures,
why true accessibility means more than formal compliance,
how the human-in-the-loop principle works in practice,
and how AI and human expertise can collaborate meaningfully.
