Election misinformation with generative AI was predicted to be catastrophic, but the observed damage varied by country. Which defenses actually worked - watermarking, platform policy, media literacy, or something unglamorous?