If they're well programmed there's no way to know at this point. AI is able to pass the Turing Test without even needing particularly hefty resources, I've been doing a bunch of fiddling with local LLMs and I could probably write something up that could do it. I don't personally see a point because I comment here for my own enjoyment rather than to push an agenda, but if I was trying to push an agenda it'd be reasonably straightforward to whip up a population of AI characters who agreed with it in various ways.