ChatGPT Can Now Control Your Chrome Tabs — and Fill Out Forms for You
OpenAI announced on October 15 that the ChatGPT Chrome extension has received a sweeping update that turns the browser into the company's primary agent delivery vehicle. With the new version, ChatGPT can read user-open tabs, compare information across different websites, and in Work mode, fill out forms, send messages, and download files directly. The extension already has 7 million active users according to the Chrome Web Store, and now represents the culmination of OpenAI's strategy following the August shutdown of its standalone Atlas browser.
From Atlas browser to Chrome extension
In July 2026, OpenAI announced it was deprecating the ChatGPT Atlas, its autonomous browser with agent mode, which stopped working on August 9. At the time, the company promised that browsing and automation capabilities would migrate to the Chrome extension and ChatGPT sidebar. With this October update, that migration reaches completion: everything Atlas did now lives inside Chrome, the browser that dominates more than 65% of the global market.
The decision makes sense from a mass adoption perspective. Convincing users to switch their default browser for an AI-native client was an uphill battle against Chrome's entrenched installed base. Meeting users where they already spend their time — browsing, shopping, researching — is a far more pragmatic strategy. The extension lets ChatGPT operate within the user's regular browsing profile, with access to accounts already logged into Gmail, LinkedIn, Salesforce, and other services.
Two modes, two audiences
The update splits capabilities into two distinct modes:
Chat mode — for casual use, the assistant can "look at" the user's open tabs to answer questions. If you have three furniture store tabs open, you can ask ChatGPT to compare dimensions, prices, and return policies without copying and pasting URLs. It's essentially a cross-tab context feature that previously required multiple text copies.
Work mode — this is the substantive upgrade. ChatGPT Work, the company's agent for complex, multi-step tasks, can now actively control Chrome tabs. Practical examples include filling out event registration forms, sending messages on web platforms, bulk downloading files, organizing tab groups, and comparing terms from multiple vendors to draft a proposal email. The work happens in the background, without "stealing" the user's active browsing session.
The crucial caveat: Work mode requires at minimum a ChatGPT Plus subscription. This positions the tool as a premium product — unlike Gemini, whose auto-browsing functionality is available free. OpenAI appears to be using browser automation as one of the primary conversion hooks for the paid tier.
What the extension can (and can't) see
The extension operates within your regular browsing profile, which means it can access any site where you're already logged in. This brings both advantages and significant privacy concerns.
According to the extension's official documentation, OpenAI does not maintain a separate complete record of browser actions. Browsing activity is stored only when it becomes part of ChatGPT's context — that is, when the assistant reads text from a page, takes screenshots, executes tool calls, or includes content in a conversation. The company emphasizes it will always ask for permission before accessing browser history or managing downloads.
In practice, this means ChatGPT can see and interact with your banking, health, work, and personal shopping tabs — as long as you allow it. OpenAI recommends closing sensitive tabs (banking, health, corporate administration) before activating Chrome in Work mode, and treating automated form-filling as an assistant whose entries should be reviewed before submission.
Removing a site from the allowlist makes ChatGPT ask for permission again. Completely removing the extension from Chrome revokes all access. But the friction point is that the removal barrier sits in the browser's extensions page — not a simple toggle within the ChatGPT conversation itself.
Sidebar, full history, and the future
Beyond tab control, the extension brought a sidebar upgrade: it now offers "the full power of ChatGPT," including the entire conversation history in Chat and Work modes. This means you can open a sidebar next to any site, ask about the page content, and receive contextual responses without ever switching tabs.
The sidebar also supports mentions of open tabs, text highlighting from pages, and questions about YouTube videos using timestamped transcripts when captions are available. Side chat features are available in Chrome, Edge, Brave, and Vivaldi — Opera doesn't support side chat but allows starting tasks in the desktop app.
Competition with Gemini Auto Browse
ChatGPT's functionality closely mirrors Google's auto-browsing Gemini feature, which recently expanded to India. Both let AI read and interact with open web pages, compare information across sites, and execute automated tasks. The main difference is the business model: Gemini is free, while ChatGPT Work requires a paid subscription.
However, ChatGPT has the advantage of a larger installed base — 7 million users on the Chrome Web Store compared to Gemini's estimated millions, with no official public numbers. And the fact that the extension works with OpenAI's most advanced model (likely GPT-6, given the October 2026 context) could be a competitive edge in reasoning quality and complex task execution.
What to watch
The extension is available to all users, free and paid. The real question is how it performs in the real world, far from the controlled demos OpenAI presents. Dynamic sites, unexpected popups, CAPTCHAs, and anti-bot checks can still trip up automated agents quickly. OpenAI has demonstrated impressive form-filling and data scraping in promotional videos, but the everyday user experience will be the true test. If the extension can reliably do what it promises, it could significantly accelerate the adoption of AI agents in daily life — while simultaneously opening a broader debate about how much autonomy we want to grant tools that operate inside our own browser.
Sources: The Verge, Unite.AI, TechRadar
✓ Independent sources cross-checked and verified before publishing
