Introduction: The use of artificial intelligence (AI) in educational assessment offers scalable solutions to traditional grading challenges, yet concerns about reliability, fairness, and acceptance remain, particularly in subjective domains like writing. This study examines undergraduate students' perceptions of AI-generated scoring and feedback compared to human evaluators.

Methods: A total of 159 undergraduates at a Canadian university reviewed standardized writing responses paired with scores and feedback from either a human evaluator or ChatGPT-4. They completed surveys assessing perceptions of AI evaluation before and after disclosure of whether the evaluator was AI or human.

Results: Students often struggled to identify the evaluator accurately. Overall, perceptions of AI scoring and feedback were moderate, but exposure to AI significantly reduced confidence in AI scoring. Both the source of grading and identification accuracy influenced perceptions. Human grading generally elicited more positive perceptions, whereas incorrect identification—when not combined with human grading—also led to more favorable perceptions. In contrast, the combination of human grading and incorrect identification resulted in more negative perceptions. Additionally, individual factors such as comfort with technology, familiarity with AI, and frequency of AI use significantly predicted students’ attitudes toward AI.

Conclusions: These findings enhance our understanding of student attitudes toward AI in educational assessment, highlighting the role of digital literacy and prior exposure in shaping perceptions. In exploring how students interpret and respond to AI-generated feedback, this study contributes to educational psychology by illustrating the cognitive factors that influence engagement with academic evaluation. The results emphasize the importance of thoughtful implementation to build trust and support acceptance of AI tools in educational contexts that foster effective learning.