OpenAI Expands ChatGPT in Google Chrome: AI Can Now Read Your Tabs and Automate Browser Tasks

 

OpenAI is bringing ChatGPT closer to the way people actually use the internet. With an expanded browser extension, ChatGPT can use information from open browser tabs, compare content across multiple websites and, through its Work mode, help carry out practical tasks such as filling in forms, sending messages, downloading files and organising tabs.

For anyone who regularly switches between research articles, shopping pages, spreadsheets and online forms, the change could make everyday work considerably more efficient. Instead of copying information from one website into a chatbot and then transferring the results elsewhere, users can ask ChatGPT to work with their existing browsing context.

The development also intensifies competition with Google Gemini, which is gaining a more prominent role in Chrome and Google's wider ecosystem. The competition is no longer just about which AI assistant provides the best answer. It is increasingly about which assistant can understand what users are doing, connect information across applications and help them complete tasks.

What's New: ChatGPT Becomes a Hands-On Browser Assistant

The headline feature is straightforward but powerful. ChatGPT can now see and interact with your open browser tabs. OpenAI framed the rollout as a way to bring the power of computer use to anyone who uses Chrome, and the capabilities back that claim up. The extension, already installed by roughly seven million users, has shifted from a sidebar chatbot into a genuine productivity layer that lives alongside your browsing session.

Key capabilities at a glance:

Cross-tab reading: ChatGPT can scan content across every open tab to answer questions, compare options, or pull specific data points.

Tab control: In Work mode, the AI can fill out forms, send messages, and download files on your behalf, with permission prompts before sensitive actions.

Tab organisation: Automatically group related tabs to cut through clutter and surface what matters.

Document creation: Generate comparison sheets, spreadsheets, and slides based on tab content.

@chrome tagging: Type @ in any ChatGPT chat to invoke the browser integration and point it at your tabs.

Side Chat: A persistent sidebar where ChatGPT can summarise YouTube videos, explain highlighted text, and reference open tabs without breaking your flow.

Page-level shortcuts: Highlight text on a page and bring it into chat, or right-click and choose 'Ask ChatGPT'. On YouTube videos with captions, ask questions using the timestamped transcript.

OpenAI demonstrated the feature with a relatable scenario: a shopper asking ChatGPT to compare three sofa listings across different tabs, pulling out dimensions, prices, delivery times, and return policies into a single, digestible answer. It's the kind of task that normally involves furious tab-switching and a notepad; now it's a single prompt. Multiply that across the hundreds of small comparisons, form-fillings, and research tasks you do every week, and the productivity upside starts to look genuinely meaningful.

Chat Mode vs. Work Mode: What's the Difference?

The update splits ChatGPT's browser powers into two distinct modes, and understanding the difference matters. The split is essentially a privacy and safety design choice; it lets users pick how much agency they want to hand over on any given task.

Chat Mode — your read-only research partner

Chat mode is the safer, read-only layer. Think of it as a supercharged research assistant. You can ask it to compare subscription plans across three tabs, summarise a stack of articles, or extract key details from a product listing, all without the AI touching anything. It reads, synthesises, and reports back. There's no risk of it accidentally clicking 'Submit' or moving data between systems, which makes it ideal for early experimentation and for tasks where you only want insight, not action.

Work Mode — the agentic layer that takes action

Work mode is where things get interesting. This is the agentic layer, where ChatGPT actively controls your browser. It can fill out registration forms using details from another tab, draft and send emails, download files in bulk, organise your tabs into groups, and enter data into business systems. Crucially, it asks for your permission before interacting with new websites, and you can instruct it to stop before completing sensitive actions like a form submission. OpenAI's own demos include filling out a new-dentist form and adding the appointment to both work and personal calendars, or comparing contractor quotes and drafting an email to the best fit.

The Side Chat feature operates alongside both modes, a persistent sidebar where you can ask about highlighted text, summarise a YouTube video currently playing, or pull context from open tabs without leaving the page you're on. It's designed for the moments when you don't want a full task delegation, just a quick contextual answer. The full ChatGPT conversation history is available there too, which means you can pick up earlier threads without breaking focus.

The mental model: Chat mode reads and explains. Work mode reads and acts. Start with Chat for trust, then graduate to Work for productivity.

Compatibility & Availability

Good news for non-Chrome users: this isn't a Chrome-only play. The extension works across Chromium-based browsers, which significantly broadens its reach, and is one of OpenAI's biggest structural advantages over Gemini, which is locked to Chrome. The full breakdown:

On the plans front, OpenAI says the upgraded extension is available across ChatGPT tiers, including free accounts. Tab-aware Chat features work for everyone; the heavier agentic actions in Work mode may see the fullest benefit on paid tiers, and exact availability can depend on rollout stage and workspace settings. If a feature is missing on your account, give it a few days; the rollout began October 15–16, 2026.

You'll need two pieces to unlock browser control: the ChatGPT desktop app (macOS or Windows), which houses the master Computer Use toggle, and the official ChatGPT Chrome extension from the Chrome Web Store. Enterprise and managed accounts may also be subject to administrator restrictions, so check with your IT team if you don't see the option.

How to Set It Up: A Step-by-Step Guide

Getting started takes about five minutes from a clean install. Here's the full setup:

1. Install the extension. Head to the Chrome Web Store and install the official ChatGPT extension. If you already have it, make sure it's updated to the latest version.

2. Get the desktop app. Download and install the ChatGPT desktop app for macOS or Windows. The browser integration is controlled from here, not from the web app.

3. Enable Computer Use. Open the desktop app, go to Settings, and toggle on the Computer Use feature. This is the master switch for browser automation. Pick your browser (expand 'More browsers' if it isn't listed) and confirm it now shows Manage.

4. Start a Work mode chat. Open a new chat and switch to Work mode (or use Codex if you have access). Chat mode is fine for read-only tasks, but Work mode is where the tab-control features shine.

5. Tag your browser. Type @ in the message box and select Chrome (or your Chromium browser) from the dropdown. You'll see a Manage confirmation prompt; accept it to grant tab access.

6. Give your command. Try something simple to start, like asking ChatGPT to look at your open tabs and compare phone prices. The AI will ask for permission before taking any actions.

7. Review before acting. Always review what ChatGPT did before submitting a form, sending a message, or completing a purchase. Permission prompts exist for a reason; read them.

A useful habit is to keep instructions specific. Rather than saying "help me with these tabs," try a concrete prompt like the one below.

Starter prompt: "Compare the three contractor websites open in my Chrome tabs and create a table covering pricing, availability, warranty terms, and customer reviews. Include source links and flag any missing information."

For file uploads inside Chrome, head to the extension's details page in your browser and turn on 'Allow access to file URLs' so ChatGPT can attach local files when needed.

Why This Matters: The Bigger Picture

The quiet end of the 'AI browser' experiment

OpenAI's retreat from Atlas and its full-court press on the Chrome extension tells us something important about where AI is actually headed. Convincing people to abandon Chrome, a browser with dominant market share, years of synced passwords, bookmarks, and extensions, was always a losing battle. The winning strategy is infiltration, not replacement. By turning the world's most popular browser into a host for its agent, OpenAI gets distribution it could never have achieved with a standalone product. Expect other AI labs to copy this playbook.

A direct shot at Google Gemini

Make no mistake: this is a competitive strike. Google has spent the last year embedding Gemini directly into Chrome at the browser level, sidebar access, tab summarisation, voice control, and the recently launched auto-browse feature that lets Gemini take actions inside the browser. It's the home-field advantage: Google owns Chrome, controls the roadmap, and ships Gemini updates as native browser features.

OpenAI's play is the opposite. Rather than build a competing browser (which it tried and abandoned), OpenAI is turning its seven-million-user Chrome extension into a parallel AI layer. That comes with real trade-offs. Gemini's native integration means it can do things Chrome extensions structurally can't: deeper system access, tighter performance, no permission prompts for routine actions. But ChatGPT's extension-based approach has a flexibility advantage Google can't easily match: it works across Edge, Brave, Opera, and Vivaldi too. For users who already live in ChatGPT for work, this update makes the assistant stickier; you don't have to context-switch between Gemini-in-Chrome and ChatGPT-in-another-tab anymore.

From chatbots to AI agents

The Chrome extension update isn't an isolated product tweak; it's part of a broader industry shift toward agentic AI. We're watching the chatbot era give way to the AI agent era, where assistants don't just answer questions but take actions on your behalf. OpenAI's computer use capabilities, Anthropic's Claude with similar browser-control features, and Microsoft's Copilot Studio all point in the same direction. The chatbot was a tool you visited; the agent is a tool that operates inside your existing workflow. Browser automation is just the most visible early battleground, because the browser is where most knowledge work actually happens.

The implications stretch further. If ChatGPT can fill out forms and download files inside Chrome today, it can probably book flights, manage invoices, and run procurement workflows tomorrow. The line between AI assistant and AI coworker is getting blurry fast, and OpenAI just took a meaningful step across it. For professionals, researchers, marketers, and small business owners, the productivity math is simple: every repetitive browser task that an agent can absorb is time returned to higher-value work.

Privacy, Security & Reliability: What You Should Know

Whenever an AI assistant gets browser access, the privacy conversation starts immediately. OpenAI has built in genuine safeguards, permission prompts before interacting with new websites, allowlists and blocklists under Settings → Computer Use → Manage, and a clear policy that the tool operates in the background without taking over your active session. Browser-history access is scoped to each request, with no always-allow setting, and OpenAI says it does not maintain a separate complete record of your browser actions. Activity is stored only when it becomes part of the chat itself.

Still, the trust gap is real, and early users should treat the integration with appropriate caution. Here's a practical checklist:

Open tabs are open books. ChatGPT sees your open tabs. Close banking dashboards, private emails, work documents, and sensitive chats before activating the integration.

Prompt injection is real. Malicious websites can contain hidden instructions designed to trick the AI into taking unintended actions. Stick to trusted sites, especially in Work mode.

Permission prompts exist for a reason. The AI will ask before interacting with new websites or downloading files. Don't rubber-stamp these; read them. Options include Allow once, Allow for this site, Allow for all sites (elevated risk), or Decline.

Start low-risk. Begin with read-only tasks in Chat mode (comparing products, summarising articles) before graduating to form-filling and file downloads in Work mode.

Verify before acting. Always review what ChatGPT did before submitting a form, sending a message, or completing a purchase. Real websites have dynamic layouts, surprise popups, and anti-bot checks that can trip agents up.

Use site-level controls. OpenAI flags "Allow for all sites" and browser-history access as elevated risk. Treat sensitive accounts, banking, healthcare, and employment with particular care.

Conclusion: The Browser Wars Enter a New Phase

OpenAI's Chrome expansion is a quietly aggressive move. By giving ChatGPT the ability to read and control browser tabs across Chrome, Edge, Brave, Opera, and Vivaldi, OpenAI has turned a seven-million-user extension into a genuine productivity layer that competes head-on with Gemini's native integration. The convenience is undeniable, the competitive stakes are enormous, and the open questions about privacy, security, and how much of our online lives we're willing to delegate will define the next phase of the AI era.

The next twelve months will be telling. Google will likely respond with deeper Gemini integration and tighter controls on what third-party extensions can do inside Chrome. OpenAI will keep pushing the agentic envelope, with browser automation as just one piece of a larger computer-use strategy. And users will have to weigh the convenience of an AI that acts against the privacy implications of letting it see everything they do online.

One thing is clear: the browser wars are no longer just about search defaults and rendering engines. They're about whose AI lives inside your tabs, and OpenAI just made sure it has a seat at the table. If you haven't updated your ChatGPT extension yet, now's the time. The web is about to start working for you. Just remember to keep a human in the loop.

Shakir Bukhari

Related Reads

• Google Chrome’s "Auto Browse" AI Agent: How Gemini 3 is Revolutionising Web Navigation /2026/01/google-chromes-auto-browse-ai-agent-how.html
  

• Your Browser's Getting a Brain: OpenAI's Game-Changing AI Browser /2025/07/your-browsers-getting-brain-openais.html
  

• OpenAI Launches New ChatGPT Chrome Extension: AI Browsing Gets Smarter Without a Separate Browser /2026/07/openai-launches-new-chatgpt-chrome.html