The problem with detecting pixels is you generally have to load the image before you find out that it's only 1x1 or fully-transparent. And by then it's too late

The problem with detecting pixels is you generally have to load the image before you find out that it's only 1x1 or fully-transparent. And by then it's too late. Apple could probably use ML to try and guess ahead of time whether a URL is a tracking pixel, but they have to load at least some of them first to be able to figure that out.