Google should just turn every webpage into an image and from there OCR it back into information. That's the only way to filter out all the crap that humans will not see.
Of course when you start taking the browser apart you can heavily optimize such process.
At some point you could even get so frustrated with existing APIs..