Web Agent

An AI system that can autonomously perform tasks on websites by interpreting user goals and executing actions such as clicking, typing, and navigating. Web agents often use large language models or multimodal models to interpret page content and determine appropriate actions. While promising for automation, web agents present accessibility concerns: most are designed for sighted interaction paradigms (mouse movements, visual screenshots), have low task success rates (often below 12%), and can be especially burdensome for screen reader users when errors occur, as recovery requires navigating between the agent interface and the webpage.