OpenAI turns ChatGPT into an agent that reads Chrome tabs and fills forms

ChatGPT can now read your open browser tabs, compare pricing, and fill out forms without you copying and pasting a single thing.

Tools · Source: Windows Report

What happened

OpenAI just released a massive update to its ChatGPT Chrome extension. The AI can now read across your open browser tabs to gather context and organize your workspace. You no longer need to copy and paste text between windows. The AI pulls everything directly from the tabs you already have open. For example, you can open three different couch listings and ask ChatGPT to compare dimensions, materials, and delivery times in one shot.

The real power unlocks in ChatGPT Work mode. This is OpenAI's agent for longer and more complex tasks. It takes actual control of your browser tabs. It can send messages, download files in bulk, and fill out registration forms using context from another tab. It stops right before hitting submit so you can review the action. It can even draft an email to a contractor based on services listed across multiple tabs.

Setup requires the ChatGPT desktop app. You install the extension, go to the Computer Use settings, and tag your browser with an at-mention in a chat. It works across Chrome, Edge, Brave, and Vivaldi. The extension operates in the background without hijacking your active browsing session. OpenAI says it will always ask for permission before referencing browser history or uploading files, and users can manage allowlists for specific domains.

Key facts

Why it matters

This changes how we build web-based workflows. Until now, AI assistants were trapped inside their own chat windows. Now, the AI sits directly inside the browser where the actual work happens. It can read authenticated sites where the user is already logged in, like Salesforce, Gmail, and internal company tools. This turns the browser into a unified execution environment. You do not need API integrations if the AI can just read the screen and click the buttons.

The second-order effect is the death of basic web scraping and form-filling scripts. If a user can just tell ChatGPT to pull vendor pricing from five open tabs and drop it into a spreadsheet, lightweight automation tools lose their edge. OpenAI is training users to rely on ChatGPT as the universal interface for the web. Builders who rely on simple data entry or cross-tab organization as their core value proposition will be wiped out by this native capability.

For builders

Basic web automation tools face obsolescence

If you charge for simple data entry or scraping tools, you are in trouble. ChatGPT can now pull data from multiple tabs and fill out forms natively. Your customers will just use their existing OpenAI subscription instead of paying you for a standalone tool.

Build for authenticated internal tools

The extension can read sites where users are already signed in. This is a massive opportunity for internal tooling. You can build workflows that assume ChatGPT will act as the bridge between your CRM and your inbox without building custom API connections.

Privacy and permission friction

OpenAI requires explicit permissions for browser history and cross-site actions. Users must choose to allow access per site. Builders need to account for this friction when designing agentic workflows that span multiple domains, as users might decline elevated risk prompts.

My take

I love how OpenAI skipped trying to force everyone into a new browser ecosystem and just hijacked Chrome instead. By letting ChatGPT read your open tabs and click around your authenticated sites, they put their agent exactly where your data already lives. This is how you win the automation war. You do not build a new destination. You take over the one people already use.

Original reporting: Windows Report. This is my rewrite and opinion.

More AI news for builders