Investigating OpenAI’s ChatGPT Capabilities to Improve Accessibility of Textual Information: An Explorative Study

With the expansion of the Internet and mobile services, digital exclusion due to accessibility barriers remains a concern, particularly in understanding complex web content. This study investigates the application of Large Language Models (LLMs) like OpenAI’s ChatGPT for simplifying sentences in line with easy-to-read (E2R) guidelines. Through two exploratory studies, we assess LLMs’ awareness of E2R guidelines and their capability to simplify sentences to enhance cognitive accessibility in an archaeological museum context. The first one aims to assess LLMs’ general knowledge about E2R guidelines. The second one evaluates their ability to perform SS conforming to E2R guidelines in order to generate accessible textual information for an archaeological museum. To this end, results of SS provided by LLMs are compared with text made by students who attended training on the E2R guidelines and those made through a co-design process with people with cognitive disabilities. Our findings support the discussion of the current potential and limitations of using LLMs as tools to improve the cognitive accessibility of text, providing insights into their potential role in making information more accessible.