Browser has ChatGPT directly integrated into the search. It promises to fulfill the traditional functions of a browser, but has an Agent Mode designed to replace the user in certain tasks.
OpenAI officially presented the ChatGPT Atlasits new internet browser that aims to transform the way users interact with the Internet by Directly integrate ChatGPT into the search experience.
The initial version, available since this Tuesday for macOS, also includes a preview of the innovative “Agent Mode”, an assistant capable of carrying out online tasks autonomously.
The announcement was made by the company’s founder and CEO, Sam Altman, who compared the launch to the impact that Google Chrome had in 2008. Atlas will allow users to “talk to a page” on the Internet, he said.
Atlas will reach Windows and mobile devices as quickly as possible, the CEO also promised.
Browser with integrated AI
With a home screen that resembles Chrome’s minimalism — a bar where the user can “Ask ChatGPT or enter a URL” — Atlas stands out, however, for the unprecedented integration of the OpenAI language model.
The browser allows you to access conversation histories, switch between different versions of ChatGPT and even receive automatic suggestions such as news or tasks that the browser can perform, explains .
OpenAI highlights that Atlas includes the traditional browser functions — such as tabs, favorites and autocomplete — but adds a decisive new feature: chat follows the user everywhere.
In other words, it is possible to search between bookmarks or history using natural language commands, open a “side chat” to ask questions about the page you are on or even edit a draft email in Gmail directly with the help of ChatGPT, without needing to switch between windows.
When the user performs a short search, Atlas responds like a language model — presenting text with embedded links to sources. But the browser also offers tabs with traditional lists of results, images, videos or news, similar to a conventional search engine.
O Agent Modecurrently only available to ChatGPT Plus and Pro subscribers, allows you to perform actions automatically based on user instructions.
Users can observe the actions of the agent in real time — which “clicks” on tabs and pages like a human user — or let it operate discreetly in the background. The mode can be activated manually or automatically suggested by ChatGPT when it detects a task that can be performed in this way.
OpenAI has ensured that this Agent Mode only has access to the interior of the browser, and that it respects user authentication and history, but without the ability to execute code outside of open tabs. The user can also choose whether each Atlas tab is “connected” or “disconnected” from other web services and use anonymous windows for private browsing.
Kill Chrome
The launch of Atlas takes place in an already growing market. Browsers like Microsoft’s Edge already integrate Copilot, while Google Chrome is preparing new functions based on Gemini, promising similar “agency” capabilities. Startups like Perplexity are also betting on AI-based browsers — and recently surprised by making a $34.5 billion proposal to buy Chrome, despite having a valuation of just $14 million.
OpenAI also expressed interest in acquiring Chrome in April, but recent legal developments in competition cases make this unlikely.
According to Reuters, which had already revealed OpenAI’s plans for a browser, the creation of Atlas represents a strategic opportunity: allowing the company to collect direct usage data and integrate advertising more naturally into the ChatGPT experience.
