Maybe those bot accounts can be detected server-side, based on their behavior.

Check how long they interact with a post before they comment.

Check if and how they zap.

Once they are suspect, add a warning icon "suspected bot", that can be removed by writing an essay judged by a human panel :)

Scroogey

Bot or not?

Signal312

Yeah. 

It's gotta be extremely difficult nowadays to algorithmically distinguish between humans and bots. Because I _know_ that some of the captchas that I run into nowadays? Some people I know, especially older ones, couldn't solve them.

I just did a little volunteer chore for a local organization, with lots of older people. It involved people clicking on a link, and filling out a simple short form. About 25% of people requested help with it.