ChatGPT Browser Extension Adds Tab Context and Control

Introduction

On October 15, 2026, the official ChatGPT account on X announced: "We're bringing the power of computer use to anyone who uses Chrome." In practice, this is an upgrade to the ChatGPT browser extension. According to The Verge, ChatGPT can now use your open Chrome tabs as context when you tag Chrome in ChatGPT's web or desktop app, and ChatGPT Work can control those tabs to carry out tasks. OpenAI's documentation page for the extension describes the setup, permissions and data handling in detail, and it is the main source for this review.

The change matters because it moves ChatGPT from answering questions about text you paste to reading and acting inside the pages you already have open, including sites where you are signed in. The release notes also give context: OpenAI said Atlas, its earlier browser, was scheduled to stop working on August 9, 2026, and that it was bringing browser-based agentic capabilities into ChatGPT and Codex. This extension is one visible form of that direction.

Feature Overview

Tab context. When you tag Chrome in the ChatGPT web or desktop app, ChatGPT can read your open tabs. The Verge's demo compared three open sofa tabs. You can also highlight text and bring it into a chat, or right-click and choose "Ask ChatGPT" in browsers that support side chat. For YouTube videos with captions, ChatGPT can use the video's timestamped transcript to explain, summarize or answer questions in a supported side chat.

Side chat. The sidebar has been upgraded with what The Verge describes as the full power of ChatGPT and your entire chat history. It can be used in Chat or Work beside the page, stays with the tab where it was opened, and chats can move between side chat and the desktop app.

Browser control. ChatGPT Work can control Chrome tabs for tasks such as sending messages, filling out forms or downloading files, and it can organize and group tabs. The Verge compares this to Google's Gemini-powered "auto browse" in Chrome. Per the docs, the extension can read or act on sites where you are already signed in, such as LinkedIn, Salesforce, Gmail and internal tools.

Browser coverage. The docs list Google Chrome, Microsoft Edge, Brave, Opera and Vivaldi, all driven from the ChatGPT desktop app. All five support tab mentions and browser control. Side chat is available in Chrome, Edge, Brave and Vivaldi, but not Opera, where tasks start in the desktop app. A separate @Browser mention uses ChatGPT's built-in browser instead of your regular profile.

Setup. Update the desktop app, open Settings > Computer Use, choose your browser, install the extension and review its permission prompts, then confirm the browser shows Manage. After that, @-mention the browser in a ChatGPT Work or Codex chat. The docs give the example "@Edge open Salesforce and update the account from these call notes."

Usability Analysis

OpenAI's sample prompts show the intended pattern: fill a registration form from an event brief and stop before submitting; put vendors from several tabs into a comparison sheet with pricing, features and source links; compare contractor tabs and then draft an email; fill a dentist intake form and add the appointment to work and personal calendars. These are multi-tab, multi-step chores where copy-and-paste is the usual friction.

The permission model is designed to keep a person in the loop. By default ChatGPT asks before interacting with each new website, per host. The choices are Allow once, Allow for this site, Allow for all sites (labelled Elevated Risk) or Decline. Allowlists and blocklists are managed in Settings > Computer Use > Manage and are shared across browsers. Browser history access is asked per request and has no always-allow option.

For day-to-day use, the practical question is how often the prompts interrupt a task. Per-host confirmation is safer, but a workflow touching many sites will prompt repeatedly, and the all-sites option removes confirmations entirely at higher risk.

Pros and Cons

On the positive side, the extension works across five Chromium-family browsers rather than one, the per-site permission prompts keep consent explicit, and the side chat connects the page you are reading to your existing chat history. Prompts that stop before submitting give a natural review point.

The limits are also documented. The Chrome install prompt can include permissions such as "Access the page debugger", "Read and change all your data on all websites" and "Read and change your browsing history on all your signed-in devices", plus downloads, bookmarks and tab groups. The docs tell users to treat page content as untrusted context, and note that malicious page content can raise the risk of data being copied somewhere unintended. Opera lacks side chat. Availability depends on rollout and workspace settings, and the sources reviewed do not state plan tiers.

Data Handling

OpenAI says it does not store a separate complete record of browser actions. What it stores is what becomes part of the chat context: page text read, screenshots and tool calls. The Memories setting applies, and standard ChatGPT data controls cover content processed in context. The docs advise avoiding secrets or highly sensitive data unless required and staying present to review each prompt.

Outlook

Both OpenAI and Google now offer a way for an assistant to operate Chrome tabs on a user's behalf, so the comparison will likely turn on reliability, permission design and how well each handles untrusted pages. The documentation's emphasis on untrusted page content suggests prompt injection remains the central risk for this category. For OpenAI, folding browser actions into ChatGPT Work and Codex chats keeps the workflow inside tools users already open.

Conclusion

The upgraded extension is a practical step for people who already spend their day across many tabs and signed-in web apps, and its per-site consent model is a reasonable baseline. It suits ChatGPT desktop users who can accept broad extension permissions and are willing to review each prompt. Teams with strict data policies should review the permission list and workspace settings before enabling it. Rating: 4 out of 5.

Editor's Verdict

ChatGPT Browser Extension Adds Tab Context and Control earns a solid recommendation within the GPT space.

The strongest case for paying attention: a single extension covers five browsers, with tab mentions and browser control in all of them. That alone raises the bar for what readers should expect in this space. Reinforcing that, per-site confirmation keeps consent explicit before ChatGPT acts on a new website — practical value rather than just headline appeal. The broader signal worth registering is straightforward: the upgrade turns the extension from a page reader into an actor on signed-in sites, which shifts the main risk from wrong answers to wrong actions. On the other side of the ledger, one constraint is real rather than a marketing footnote: the Chrome install prompt can request broad permissions, including reading and changing all data on all websites and the page debugger. It should factor into any serious decision. Layered on top of that, malicious page content can raise the risk of data being copied somewhere unintended, per OpenAI's own safety guidance — which narrows the set of teams for whom this is an obvious yes.

For ChatGPT desktop users who work across many signed-in web apps, and for workspace admins weighing browser-agent permissions, this is a serious evaluation candidate, not just a curiosity to bookmark. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.