AI DC – Artificial intelligence supporting digital accessibility

Project factsheet

Project summary

Researchers from the Institute of Computer Science, Polish Academy of Sciences and Sages, commissioned by the Ministry of Digital Affairs, prepared an analysis of the possibilities of using artificial intelligence in digital accessibility research. The analysis was carried out within the project “AI DC – artificial intelligence supporting digital accessibility”, implemented by the Ministry under the European Funds for Social Development 2021–2027 (FERS) programme. The aim of the Ministry’s project is to create an AI-based virtual assistant that will help creators of digital solutions in Poland implement digital accessibility.

The task of the analysis was to identify the technologies with the best potential for testing and ensuring digital accessibility, including a review of the language models available on the AI market, with emphasis on open-source solutions. It reflects the situation on the AI market as of the end of 2025.

Results

The report “Analiza możliwości wykorzystania sztucznej inteligencji w obszarze badania dostępności cyfrowej” (Analysis of the possibilities of using artificial intelligence in digital accessibility research), dated December 2025, was published by the Ministry of Digital Affairs on the GOV.pl website under the CC BY 4.0 licence. The report team consisted of Aleksandra Charyło, Ariel Drozd, Jakub Konieczny, Łukasz Kobyliński, Michał Marcińczuk, Aleksandra Zwierzchowska and Damian Żłobicki.

The report:

• introduces the WCAG guidelines, the characteristics of various types of digital content and the current state of automated accessibility audits, pointing out their limitations and the potential for further automation with AI;
• describes the market of large language models and LLM-based systems, together with the main challenges related to this class of models;
• analyses the potential of AI models for specific accessibility tasks: analysis of HTML, JavaScript and CSS code, summarisation and simplification of content, proofreading, machine translation, alternative text, audio description, speech synthesis, audio and video transcription, document layout analysis, optical character recognition (OCR) and text/background contrast analysis – for many tasks the results of the best models are promising, although not all of them carry over to Polish-language data;
• recommends investing public funds in dedicated tools based on open technologies, and documenting reported accessibility violations to set priorities and validate the proposed solutions;
• reviews the technologies recommended for building the target tool, proposes sets of models based on the state of the art in 2025 and presents a proof of concept of an AI-based system for processing PDF files;
• discusses the potential for further training of base models and forecasts the development of AI in the context of digital accessibility.

See also the news item: Report on the Use of AI in Digital Accessibility Research.