Filtering out bots is prohibitive, as bots are currently so close to human text that the false positive rate will curtail human participation.

Any community that ends up creating utility to its users, will attract automation, as someone tries to extract, or even destroy that utility.

A potential option could be figuring out community rules that ensure all content. including bot generated content, provides utility to users. Something like the rules on change my view, or r/AITA. Theres also tests being run to see if LLMs can identify or provide bridges across flamewars.