How GPTs see the Internet

Chris Long highlighted the main point of Dan Petrovich's articles:

AI-based robots do not look at the entire page. Instead, they analyze small «windows» of text.
The only content that the robot is guaranteed to extract is the title, URL, and text snippets (sometimes extracted from the meta description).
As for the content of the page, it doesn't view all the HTML code automatically. Instead, it can view the content in different «sliding windows». The open() function is used to open on different lines, simulating page scrolling.
In theory, AI-powered search could prioritize the top of the page. That's why it's important to place key content here, as it's more likely to be placed in the window.