How can the increasing dependence on AI-generated answers affect the integrity of online knowledge, especially when AI mistakes or speculations get repeated and ultimately solidify into accepted facts? What solutions could be implemented to prevent this cycle and ensure factual reliability and avoid biase?