This study examines how artificial intelligence, specifically the large language model (LLM) ChatGPT-4o, frames religious content in comparison to Swedish religious education textbooks. As generative AI becomes increasingly common among students, including for schoolwork, it is important to understand how these tools present religious and ethical topics. Using a qualitative comparative approach and an adjusted version of Entman’s framing theory, this study analyzes how five Swedish religious educational textbooks and ChatGPT-4o frame religious knowledge across five world religions: Christianity, Islam, Hinduism, Buddhism and Judaism. For each of these religions, the study chose a single Swedish religious educational textbook and a single topic. The study then compared and contrasted the textbook’s and ChatGPT-4o’s presentations of these topics. The results reveal that while ChatGPT-4o offers structured and neutral summaries, it often avoids moral perspectives, critical reflection, and deeper engagement with complex topics. This is especially clear in ethically sensitive topics such as discrimination or gender equality, where the textbooks provide deeper insights, more diverse voices and encourage reflection. This study concludes that LLMs may be useful for learning factual topics, but they are not suitable when dealing with more nuanced and complex topics. Hence, ChatGPT-4o is not appropriate as a standalone educational tool in religious education. Instead, LLMs like ChatGPT-4o should be used with teacher guidance, alongside traditional materials, to support critical thinking and discussion in the classroom.