Masterarbeit / Bachelorarbeit
Optimizing Web UIs for AI browsers
Research Area
Advisers
Dr.-Ing. Sebastian Heil
Description
The Web is facing major changes in the way that users interact with its resources. While for many years graphical web user interfaces rendered in a browser were the main mechanism to access and navigate the vast information space of the Web and to perform actions in interactive web applications, the advent of Agentic AI seems to disrupt established patterns. A new generation of browsers like Comet natively integrate AI to automate parts of the interaction with web systems. Thus, web user interfaces change their role towards becoming interfaces for both human and machine interaction. This raises the question how future web user interfaces can be optimized towards AI browsers without jeopardizing their user-centric support for human users.
This thesis aims at creating a method for optimizing existing web user interfaces for AI browsers. The method needs to be based on empirically grounded best practices and supported by software infrastructure that helps developers to refactor their web user interfaces. This especially requires comparative experimentation with different versions of web user interfaces and objective measurements of the performance improvements, for instance with regard to findability of information, accuracy of AI-generated answers, number of requests needed and overall time to answer a user's request. The optimization guidelines need to cover different types of information and representation such as information from textual content, dates and rooms, or links to relevant resources represented in text paragraphs, tables etc.
The objective of this thesis is the creation of a solution or the combination of existing approaches to solve the problem of optimizing web user interfaces based on the characteristics of AI browsers through a method and infrastructure as described above. This comprises the analysis of the state of the art of methods and techniques for optimizing web user interfaces and content representations as well as the demonstration of the solution by prototypical implementation and a suitable evaluation based on applying the method to an existing web application such as the homepage of the professorship.
