A floating cross-browser extension that brings instant AI actions — summarise, ask, simplify, and read aloud — to any webpage without leaving the site you're on.
Problem
Switching between browser tabs to summarise an article, look up an answer, or simplify dense jargon is a constant friction point in everyday research and reading. Browser-native tools offer nothing; general AI assistants require leaving the page entirely. And for users who rely on assistive technology, jumping between tools breaks focus and accessibility in equal measure. There is no lightweight, always-available AI layer that works on top of any website without disrupting the reading or working experience.
Overview
Assistive Touch AI Widget is a cross-browser extension that injects a small floating button onto every webpage. When triggered, it provides four assistive actions directly on the current page: summarise the content, ask a question about it, simplify dense or jargon-heavy text into plain English, or read the selection aloud using the browser's built-in speech engine. The widget reads selected text first, falling back to full-page content when nothing is selected. AI actions run through the user's own API key, stored locally in the browser — no central server holds credentials.
How It Works (Approach)
The extension is built with Manifest V3 and uses the WebExtensions polyfill to run on Chrome, Edge, and Firefox from a single codebase. A content script injects a React component into the page via shadow DOM, isolating the widget's styles from the host page entirely. AI actions call OpenRouter directly using the user's stored key, while the read-aloud action uses the browser's native Web Speech API at zero cost. Results return in under two seconds and display inline in the floating panel. The widget is draggable and persists across page interactions without reloading.
Impact / Value
The widget removes the tab-switching overhead that interrupts reading and research workflows. Because it works on any page without modifying the host site, it functions as a lightweight universal AI layer for the web — summarising dense articles, simplifying jargon, answering specific questions, and offering audio playback for accessibility, all without the user ever leaving the page.
Key Features
- Floating draggable button on every webpage
- Summarise: 3–5 bullet points from selected text or full page
- Ask: any question about the current page content
- Simplify: rewrite dense text in plain, accessible English
- Read aloud: built-in text-to-speech via the Web Speech API
- Shadow DOM injection — zero style conflicts with host pages
- Cross-browser: Chrome, Edge, and Firefox from one codebase
- Bring-your-own API key — no central server, no shared credentials