A problem is that an AI, when it interrogates forums, does not know fact from fiction. Consequently, it has to weigh each "interrogation" before generating its response. Hopefully, the underlying algorithm will "pick" the best response.
PS: I also noticed that an AI may not be time sensitive...