Some of the manipulation seems obvious. Some of it less so. I always felt like some kind ublock origin style plugin that highlights or removes known instances of PR/manipulation/propanda would be helpful. Perhaps an AI trained on press releases from big corps and governments that performs authorship identification for a start. 'The writing style in this post are an 80% match for a press release by <government body> linked below'. Kind of like those websites that try to identify fake product reviews.

Theres a high chance that would probably get gamed, too, I guess.