How transparent are the sources AI draws from? Testing chatbots’ resilience to disinformation