Constructing a Generative AI Framework for Personalised Instructional Resource Development: A Quasi-Experimental Evaluation in a Micro-Lecture Design and Production Course

Generative artificial intelligence is promoted as a way of tailoring instruction to learners, yet most reported applications stay within language and general-education settings, and few measure outcomes in courses whose value lies in design and production skill. This paper describes a four-module generative AI framework for resource development that links learner and content modelling, generation, assessment and community iteration to a course-grounded knowledge base, with the teacher reviewing before release. It was evaluated in a twelve-week quasi-experiment in a professional core course, Micro-lecture Design and Production, at a Chinese vocational college. Two intact classes (n = 45, 44) were taught by the same instructor, who designed the framework; assignment was not random. Covariance-adjusted posttest means differed by 5.6 points (82.3 versus 76.7, p < .001, partial η² = .24), and the framework-supported class reported higher motivation and satisfaction; self-directed learning differed modestly. Preparation time per unit fell by half, though the work became more concentrated and the knowledge base carried a large one-off cost. Because the design cannot separate the framework from the additional resources it produced, the results are framed as associations rather than effects.