WebPerceptor: An Open Source Chromium Plugin for Real-Time LLM-Based In-Line, In-Browser Re-Writing of Website Content

Read full text
Authors: Joseph O’Hagan, Graham Wilson, Mark McGill
Cited by: 3
References: 17

Abstract

We introduce the "AI Mediated Web", an automatic, personalised, in-line, real-time remixing of web content by an LLM based on given prompts and presented in-browser such that the user perceives the end result as a published web page. We present WebPerceptor an open source, client-side Chromium plugin which, for any webpage, identifies text content, relays this to a local or cloud-based LLM with a user-defined prompt and then modifies the existing text (to replace, append, or remove it) based on the LLM response. We discuss the potential benefits and harms of an AI mediated web where the idea of a shared, canonical, objective web will no longer hold as AI becomes the mediator of what we perceive when we browse online.

Fields of Study & Subjects

Web Data Mining and Analysis Digital Humanities and Scholarship Web Application Security Vulnerabilities

Publication details

Journal / source
Proceedings of the Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems
Publisher
ACM
Pages
1–5
Publication date
13 April 2026
Language
EN