Misinformation spreads rapidly and at a large scale on social media, with consequences for public opinion formation and democracy. Platforms therefore increasingly rely on artificial intelligence (AI) to correct misleading content cost-effectively and on a broad scale. However, the form in which these corrections should be offered remains an open question. A static fact-check under a post reaches everyone who sees it, but delivers the same information to all. In contrast, an AI chatbot can address individual questions, but reaches only those who choose to use it voluntarily. This project investigates how both formats work. Do they help distinguish misinformation from accurate posts, and do they reduce the willingness to share misinformation? Do the formats complement each other? And does it matter whether a correction is labeled as AI-generated or written by experts? We address these questions using online experiments in a simulated social media environment.