The rise of ChatGPT has sparked interest among educators in integrating it into teaching and learning. However, effective teaching requires a deep understanding that allows instructors to use multiple representations to support comprehension and address students’ misconceptions. Before relying on ChatGPT as a teaching tool, it is crucial to assess its ability to interpret multiple representations. This study evaluates ChatGPT’s understanding of the fundamental statistical concept “data = model + error,” which underpins a number of statistical analyses in introductory statistics courses. Through tasks involving graphical representations, we qualitatively examined GPT models’ understanding and compared their strengths and limitations to those of students. The results showed that while ChatGPT demonstrated competence in certain areas, it also exhibited misunderstandings—some resembling students’ and others unique to the models themselves.