As robots become more sophisticated and human-like, how can we be sure what they tell us is real?