OpenAI Gives ChatGPT Eyes on Your Chrome Tabs
OpenAI has upgraded its ChatGPT extension for Google Chrome so the assistant can read and reason over tabs a user already has open The Verge. With the extension installed, users tag Chrome inside ChatGPT's web or desktop app to give it visibility into those tabs.
The example OpenAI gave asks ChatGPT to compare three open tabs showing sofas under consideration. Context comes directly from the pages named in the prompt.
The update also changes ChatGPT in the Chrome sidebar. It now provides full ChatGPT power and entire conversation history in Chat or Work without leaving the page being viewed The Verge. Work can control Chrome tabs for tasks including sending messages, filling out forms, or downloading files. The upgraded extension can also organize or group open tabs.
Beyond Chrome, documentation describes the browser extension as able to work in an existing Chrome, Edge, Brave, Opera, or Vivaldi tab, and as the path used when ChatGPT needs to use the user's regular browser profile OpenAI developer documentation. All five supported browsers support tab mentions and browser control from the ChatGPT desktop app OpenAI developer documentation. In the desktop app's built-in browser, user and ChatGPT can see the same page, and ChatGPT can work across tabs and download files OpenAI help documentation.
The tab work follows earlier browser steps. OpenAI previously offered a Chrome extension to search directly via the browser URL bar OpenAI. In July 2025, Reuters reported that OpenAI planned to release a web browser to challenge Google Chrome Reuters, designed to keep some user interactions within a ChatGPT-like native chat interface instead of clicking through to websites.
The broader context here is a move from isolated chat to browser-native assistance. Tagging tabs replaces manual copying of links and text with live page context. Sidebar persistence means history and model capability travel across navigation. Profile reuse matters because authentication and session state stay intact.
In this author's view, the split between Chat and Work deserves attention from practitioners. Read access to tabs simplifies summarization and comparison. Control that can message, submit forms, and fetch files introduces permission and error-handling questions. Worth flagging for enterprise use is scoping. Tab grouping and explicit mentions offer a basic control plane for what the model may see and touch. That control plane will need the same care as file permissions once assistants operate routinely across many open pages. If that holds, comparison, triage, and form-heavy workflows get less tedious while users stay where the work already lives.
