Problem-solving ability is an important skill that students need to develop in mathematics learning; however, this ability is still relatively low, especially in the topic of Systems of Linear Equations in Two Variables (SLETV). The development of Artificial Intelligence, particularly ChatGPT, provides opportunities to support students’ mathematical problem-solving processes. This study aimed to describe junior high school students’ mathematical problem-solving abilities in algebra by utilizing Artificial Intelligence based on high, moderate, and low mathematical ability levels. This study employed a qualitative descriptive approach with a case study design. The research subjects consisted of three seventh-grade students at a junior high school in Surabaya selected based on the results of a mathematics ability test and an AI usage questionnaire. Data were collected through mathematical problem-solving tests completed with and without utilizing ChatGPT, as well as interviews. Data analysis was conducted through data reduction, data presentation, and conclusion drawing based on Schoenfeld’s problem-solving stages, namely reading and analysis, exploration, planning, implementation, and verification and checking. The findings showed that students with high mathematical ability utilized ChatGPT as a reference to obtain problem-solving strategies while still relying on their own understanding during the solving and verification processes. Students with moderate mathematical ability utilized ChatGPT as guidance in determining solution steps, whereas students with low mathematical ability utilized ChatGPT to obtain solution procedures used to solve the problems. The findings indicate that the role of ChatGPT differs across students’ mathematical ability levels.