AI-generated disinformation will soon outpace our ability to detect it. What happens when falsehoods look more convincing than the truth?