Yeah, some sites (like Amazon) can really be awful in some ways. I'd certainly use LLMs to "de-enshittify" UIs, fighting fire with fire.

Making meaningful decisions though? No way.