I believe that there is no perfect solution to reduce bots. Whatever a filter defines as undesirable, bots can be tweaked to avoid triggering those characteristics.

However, if we assume that price or 'trading advice' bots are deployed to impact the market or in some way generate revenue, then filters can be designed to reduce that type of activity. I think that's a category specific enough to be separated linguistically from normal conversations. How bots will be tweaked to respond to such a filter remains to be seen.