Debunking with Dialogue? Exploring AI-Generated Counterspeech to Challenge Conspiracy Theories