Atlas, OpenAI's new ChatGPT-powered browser, launched Tuesday with the aim of "rethinking what a browser can do" and challenging Google Chrome's dominance. But how accurate are AI browsers?

US artificial intelligence company OpenAI said Tuesday it is launching its own web browser, Atlas, to rival Google 's popular Chrome browser .

Atlas will be powered by OpenAI's popular ChatGPT chatbot , as the California-based company seeks to revolutionize the way people use the internet.

“Tabs were great, but we haven’t seen much browser innovation since then,” OpenAI CEO Sam Altman said in a video presentation released Tuesday, calling it a “rare, once-in-a-decade opportunity to rethink what a browser can be and how to use one.”

What is different from ChatGPT Atlas?

For example, Altman suggested that the classic URL search bar at the heart of traditional browsers could be replaced by an AI chatbot interface.

The browser will initially only be available for Apple's Mac computers, the company said, adding that it is designed to help users perform "tasks without copying and pasting or navigating away from the page."

Another key feature of the Atlas browser is its so-called "agent mode," which effectively surfs the internet automatically on behalf of the user, armed with a person's browser history and predicting the type of information they are likely to search for.

"It's using the internet for you," Altman said.

Review of AI-powered browsers

That's one way to look at it. But analyst Paddy Harrington of London- based market research group Forrester warned that another way to think about OpenAI's new browser is that it "takes away your personality."

"Your profile will be personally tuned to you based on all the information that's been sucked up about you," Harrington told the Associated Press (AP). "OK, scary. But is it really you, really what you think, or what this engine decides it's going to do? And will it add preferred solutions [to user queries] based on ads?"

Regardless, Harrington said it will be a big challenge for Atlas to "compete with a giant that has a ridiculous market share."

Atlas launched in the market dominated by Chrome

Since its launch in 2008, Google Chrome has amassed approximately 3 billion users worldwide, blowing away rivals such as Microsoft's Internet Explorer and later Edge browsers.

But AI chatbots such as ChatGPT are increasingly summarizing information from the internet so effectively that many users are turning to them instead of the traditional practice of clicking on links suggested by a browser.

OpenAI said ChatGPT already has more than 800 million users, while a survey conducted on behalf of AP this year found that about 60% of Americans—and 74% of those under 30—use AI to find information at least sometimes.

Browsers such as Chrome have also integrated AI summaries into their search results, typically visible at the top of the results page above the first link, although concerns have been raised about the accuracy of this information.

How accurate is the information from AI?

On Wednesday, research published by the European Broadcasting Union (EBU) and the BBC found that 45% of AI responses studied contained at least one significant problem, with 81% having some form of problem.

The study evaluated AI assistants in 14 languages for accuracy, sourcing, and ability to distinguish opinion from fact, including ChatGPT, Copilot, Gemini, and Perplexity.

In one instance, ChatGPT was found to be confidently reporting that the current pope is Pope Francis, several months after the former pontiff's death—the kind of error that can significantly undermine trust and reliability, with potentially harmful effects, according to EBU media director Jean Philip De Tender.

"When people don't know what to trust, they end up trusting nothing at all, and that can discourage democratic participation," he told the DPA news agency.