Purpose: This thesis examines how university students form, revise, and limit trust in ChatGPT as an AI based digital service used for academic purposes. While previous research often links trust in digital technologies to adoption and continued use, ChatGPT presents a more complex case because students may continue to use the service even when their trust is incomplete, conditional, or task dependent. Therefore, the purpose of this study is to develop an interpretive understanding of trust formation in relation to perceived value, verification practices, and boundaries of reliance.

Methodology: The study is based on a qualitative research design using interviews with ten university students who use the paid version of ChatGPT for academic work. The interview material was analysed through thematic analysis. This approach made it possible to identify patterns in how students describe usefulness, reliability, verification, academic responsibility, and appropriate reliance on ChatGPT.

Findings: The findings show that students form trust in ChatGPT when it provides value in use, such as saving time, explaining difficult concepts, supporting writing, generating ideas, and assisting with technical tasks. However, this trust is not unconditional. Students revise and limit their reliance through source checking, comparison with other tools, repeated testing, prompt refinement, and their own academic judgement.

Conclusion: The thesis concludes that trust in ChatGPT is best understood as controlled reliance. Students continue to use ChatGPT because it is useful, accessible, and flexible, but they also maintain boundaries around accuracy, academic integrity, privacy, and personal responsibility. The study contributes to Business Administration research by showing that trust in AI based digital services is not simply a stable attitude or adoption factor, but a negotiated process shaped through use, verification, and boundary setting practices.