The use of Artificial Intelligence (AI) chatbots is becoming increasingly popular worldwide. To prevent potential negative consequences associated with the use at an early stage, it is crucial to understand its associations with mental health. Therefore, the present study compared people who use AI chatbots with those who do not engage in AI chatbot-interactions in terms of mental health and loneliness. Moreover, we investigated potential mechanisms associated with addictive AI chatbot use. Data of 7,312 persons (3,729 users: 42.1% men, 57.9% women; Mage = 42.23, SDage = 13.33; 3,583 non-users: 42.6% men, 57.4% women; Mage, = 54.28, SDage = 14.49) from the German adult population were collected via an online survey. AI chatbot users had significantly higher levels of loneliness, depressive, anxiety and stress symptoms, and significantly lower levels of life satisfaction than non-users. Users primarily engaged in AI chatbot-interactions for informational and practical reasons, such as obtaining information and inspiration, and receiving text-related assistance. The frequency of AI chatbot use to escape negative emotions was significantly positively associated with addictive tendencies. Loneliness statistically mediated the association between addictive AI chatbot use and the mental health variables. Most observed effect sizes were small. The present findings show that using AI chatbots for emotion regulation is positively associated with higher addictive tendencies and lower levels of mental health. Further experimental longitudinal research is urgently needed to clarify causal mechanisms of these associations.