Ironically that's an AI tool I would use - one that can dynamically filter content from sites according to my preferences, to counter the algorithmic spam. It wouldn't be generative AI though, and that's the only kind of AI that matters right now apparently.
Could be a good use of structured output, Llms work okay as one shot classifiers, you could define a schema that's just like, array[]{xpath:true/false} and tell the bot what you want to see and what you don't want to see.