OpenAI's ChatGPT Chrome extension can now read and act on your open browser tabs

OpenAI's ChatGPT assistant can now read the tabs you have open in Google Chrome. According to The Verge, the capability arrives through an update to the ChatGPT Chrome extension, and it lands after OpenAI retired its standalone ChatGPT Atlas browser — the company is now delivering its assistant inside the browser most people already use.

How the tab-reading feature works

With the extension installed, users can tag Chrome while chatting in the ChatGPT web or desktop app, and the assistant gains visibility into their open tabs, The Verge reports. OpenAI demonstrated the feature with a video showing a shopper asking ChatGPT to compare three tabs of sofas under consideration.

OpenAI also published sample prompts that illustrate the intended use cases. They include pulling vendor details from multiple tabs into a comparison sheet with pricing, key features and source links; comparing contractor services across tabs and drafting an email to whichever one best fits a home renovation; using an event brief in one tab to complete a registration form in another while stopping short of submitting it; and filling out a dentist intake form and then adding the appointment to both a work and a personal calendar.

The assistant can also take control of tabs

Beyond reading, The Verge describes two related updates. The ChatGPT sidebar in Chrome now carries the full assistant experience, including a user's complete conversation history from Chat or Work, so they no longer have to leave the page they are viewing.

Separately, ChatGPT Work users can let the assistant operate Chrome directly. Reported tasks include sending messages, filling in forms and downloading files, as well as tidying up a session by organizing or grouping open tabs. The Verge likens this to Google's Gemini-powered "auto browse" capability, which automates similar actions on the user's behalf.

Why it matters

Two things stand out in this update. First, it moves the assistant from a separate destination to the place where much of everyday work already happens. Comparing purchases, merging research into a spreadsheet or completing repetitive forms usually means juggling a dozen tabs; letting a model act on that live context removes the copy-and-paste middle step that has defined most chatbot workflows so far.

Second, it signals a strategic pivot. Instead of continuing to push a dedicated browser in Atlas, OpenAI is embedding ChatGPT into Chrome — Google's home turf — just as Google builds comparable automation into its own products. The rivalry between the two companies now extends to who gets to automate the browsing session itself.

There is a trade-off to watch. Giving a model sight of every open tab, including sensitive ones, plus the ability to send messages and submit forms is a substantial grant of trust. OpenAI's example prompts suggest some restraint, such as pausing before a form is submitted, but how much oversight users actually retain in practice will determine whether this convenience reads as a helper or a risk.