How to Download the ChatGPT Atlas Browser 

As of November 23, 2025, when I discovered Atlas, the Windows version has not been released yet, and the macOS version is available at:

https://chatgpt.com/en/atlas/

Just click Download for macOS there.

Short Impressions 

Whew... My heart is racing. I’m excited. This feels like a whole new world.

I haven’t tested everything yet, but when you open the browser, there is a “Ask ChatGPT” button at the top right. When you click it, a ChatGPT sidebar appears, and now you can talk directly with ChatGPT about whatever is on your browser window.

I tried it by visiting my blog first. 

Screenshot 2025-11-23 4.47.56 PM
▲ Asking ChatGPT to “explain the website I’m currently on.”

Ugh… I said, “Explain the website I’m currently on,” and it really explained everything in detail. 

At that moment, I thought, “Okay, this is insane.” But then I asked one more thing.

“Can you click the button at the top of the screen for me?”

Screenshot 2025-11-23 4.49.30 PM
▲ Using the Agent feature to actually click the menu button.

It literally clicked the menu using the Agent feature… Isn’t that insane? Truly, unbelievably insane.

For blind and visually impaired users, is there anything more revolutionary than this? Honestly? Are screen readers going to become outdated?

I still haven’t gotten over the shock.

Okay… let me calm down again.

The possibilities are endless. And here’s why:

A browser sends a request to the server, receives the document, and renders it on the screen, right?

Which means ChatGPT can directly access and analyze the actual client-side code on my machine. It doesn’t need to capture the screen, analyze images, or extract text manually. 

So I tested it again. 

Screenshot 2025-11-23 5.04.06 PM
▲ Asking about DOM structure inside the browser.

“Do you also have access to the DOM structure of this website? Since you’re a browser agent?”

“Yes, the browser agent can access the current page’s DOM structure to identify elements and use it for clicking or scrolling. So if you ask me to click a specific button or enter text somewhere, I can analyze the DOM and perform the action. I can’t show you the entire DOM directly, but I can locate and manipulate specific elements when needed.”

It’s incredible… absolutely incredible. 

Thanks to this kind of technology, I’m no longer afraid of losing my sight. I’m really living in a good era.