TLDR
- OpenAI launched Atlas, an AI-powered browser built around ChatGPT, currently available only on macOS for paying users
- The browser features “agent mode” that can research, automate tasks, and complete online shopping by opening tabs and clicking through websites on behalf of users
- Atlas can remember context from sites users visit and includes safety features like pausing before actions on sensitive sites such as financial institutions
- The browser directly challenges Google Chrome’s 71.9% market share in the global browser market
- Atlas is available globally on macOS with Windows, iOS and Android versions planned for later release
OpenAI unveiled ChatGPT Atlas on Tuesday, an AI-powered web browser that lets users automate tasks and interact with websites through artificial intelligence. The browser is currently available globally on macOS for paying ChatGPT users.
Meet our new browserâChatGPT Atlas.
Available today on macOS: https://t.co/UFKSQXvwHT pic.twitter.com/AakZyUk2BV
— OpenAI (@OpenAI) October 21, 2025
Atlas represents OpenAI’s entry into the browser market, which Google Chrome dominates with 71.9% market share. The company aims to capitalize on its 800 million weekly active ChatGPT users by expanding into more aspects of their online lives.
The browser includes several features designed around ChatGPT integration. Users can open a ChatGPT sidebar in any window to summarize content, compare products or analyze data from websites. The browser also has memory capabilities that let ChatGPT remember context from sites users visit.
Agent Mode Handles Complex Tasks
The standout feature is “agent mode,” which allows ChatGPT to interact with websites on behalf of users. In this mode, the AI can complete tasks from start to finish without constant user input. OpenAI demonstrated this by showing how the agent could find an online recipe and automatically purchase all ingredients from Instacart, navigating the website and adding items to the cart.
The agent can open tabs, click through pages, research information, and book appointments while users continue browsing. This automation takes several minutes to complete depending on task complexity. Users can also ask the agent to organize events like dinner parties by finding recipes, locating stores with ingredients, and arranging delivery.
Atlas joins a growing field of AI browsers that includes Perplexity’s Comet, Brave Browser and Opera’s Neon. These browsers compete by offering tools that summarize pages, fill out forms and draft code to attract users away from traditional options.
Safety Features and Privacy Controls
OpenAI built several safety measures into Atlas. The browser cannot run code, download files, or install extensions without permission. It cannot access other apps on users’ computers or their file systems.
The browser pauses before taking actions on sensitive sites such as financial institutions. This gives users control over what the AI agent can access and modify on their behalf.
Alphabet shares, which owns Chrome, fell 1.8% in afternoon trading following the Atlas announcement. Analysts view the browser as a potential precursor to OpenAI entering the advertising market, which could compete with Google’s search advertising business.
Google has adapted to AI-driven search behavior since ChatGPT’s launch in late 2022. The company now displays AI overviews alongside traditional links in search results. Last month, Google integrated its Gemini AI model into Chrome for U.S. users, with plans to expand to the iOS Chrome app.
OpenAI plans to release Windows, iOS and Android versions of Atlas later. The browser is part of the company’s broader strategy to compete with Google and startup Anthropic by finding new areas of growth beyond its chatbot.
