WCAG has been around for several years; however, there are still many inaccessible websites on the web. Despite the many laws and standards around the world, many developers thought that adding accessible measures to websites made them less attractive and was only favorable to a small minority. There were few demands from the majority of customers to make websites accessible. Therefore, many ignored the existing guidelines even though it would create a larger audience. AI recently became popular with the release of ChatGPT.

Since its release, people have been trying to find more ways to use AI to help with their work. AI has been used in several ways to help with accessibility, from image recognition, alt-text generators, and video captioning to intelligent virtual assistants, sign language recognition, and web design features. One aspect where AI could help with accessibility was through evaluation.

This study aimed to answer what variations exist among AI tools designed for web accessibility evaluation, and how do these tools differ in their implementation of AI? A comparative study was conducted where multiple AI tools for accessibility assessment were compared and tested on their ability to find accessibility violations on websites and, hopefully, also give recommendations on how to fix them. A self-created GPT was included among the AI tools. The websites tested were the largest public universities by enrollment websites around the world. Two websites per continent were selected and tested in each selected AI tool.

The results were then analyzed and discussed against the Universal Design Theory and previous research to shed some light on the answers to the research question. In conclusion, the study found that existing AI tools for web accessibility evaluation varied in reliability and comprehensiveness, highlighting the need for continued human involvement to ensure accurate and inclusive accessibility assessments.