OpenAI is already killing the browser it launched last October
Signals InboxJuly 10, 2026Consumer AI
OpenAI is shutting down ChatGPT Atlas less than ten months after launching it. The signal? Even the company behind ChatGPT could not convince people to leave Chrome.
The ChatGPT browser is already dead https://t.co/XR1LGI1H0f
— The Verge (@verge) July 9, 2026
Failed products by OpenAI:
— Katie Miller (@KatieMiller) July 10, 2026
🚩 ChatGPT’s Browser Atlas
🚩 Sora Video Generator
🚩 Three Stargate projects
🚩 In-app shopping with direct checkout.
🚩Hardware device
It’s almost like Sam Altman continually gaslights his investors. https://t.co/2wOkyzJuAj
Q1What is OpenAI officially doing?
OpenAI is deprecating ChatGPT Atlas, with the browser scheduled to stop working on August 9, 2026. Atlas launched on macOS in October 2025, so it lasted less than ten months. OpenAI says it is moving its browser-based agent capabilities into ChatGPT instead.
Q2Was Atlas supposed to compete with Chrome?
Yes. Atlas was not just a ChatGPT sidebar but a full browser where ChatGPT could understand open pages, remember browsing context and perform tasks across websites. OpenAI wanted the browser itself to become an AI assistant. The problem was convincing people that these features were worth leaving Chrome for.
Q3Why is Chrome so difficult to kill?
Because a browser is mostly a habit and distribution game. Chrome already holds roughly two-thirds of worldwide browser usage and an even larger share on desktop in some datasets. It stores people’s passwords, bookmarks, extensions, payment details and years of browsing history. A new browser cannot just be slightly smarter. It has to be much better.
Q4What did OpenAI get wrong with Atlas?
Atlas gave people interesting AI features, but not enough reasons to move their entire digital life. It also launched only on macOS, while Chrome works almost everywhere. Users could already access ChatGPT from Chrome, so switching browsers often felt unnecessary. OpenAI built a new front door, but most people were comfortable using ChatGPT through the old one.
Q5Did the agent features actually work?
They worked well enough to show the idea, but not reliably enough to replace normal browsing. AI agents can still misread pages, click the wrong buttons, get blocked by logins or become confused by changing website layouts. Giving an agent access to emails, accounts and payments also creates security risks that normal browsers do not have in the same way.
Q6Are other AI browsers doing any better?
The race is not over. Perplexity has Comet, The Browser Company has Dia, Opera has Neon and Brave has added Leo. Google is also putting Gemini directly inside Chrome. That may be the biggest problem for challengers: Google can copy the most useful AI features without asking billions of Chrome users to change browsers.