<p>The rapid expansion of artificial intelligence has increased the need for graduates who possess practical competence in machine learning (ML), particularly in developing countries where instructional resources are limited. However, many computer science education students in Nigerian universities complete their programmes without adequate hands-on ML skills. This study examined the effect of AI-based instruction, supported by ChatGPT and Copilot, compared with conventional video-based instruction on the ML competencies of computer and robotics undergraduates’ education. A quasi-experimental non-equivalent pretest–posttest control design was adopted, involving 61 students enrolled in a compulsory Introduction to Machine Learning course. Participants were assigned to either an AI-based project-driven instructional group (<i>n</i> = 25) or a video-based instruction group (<i>n</i> = 36). Over 10&#xa0;weeks, both groups covered identical ML content, but differed in instructional approach. Data were collected using validated instruments measuring data modelling, data preprocessing, data model construction, and data evaluation. MANCOVA and ANCOVA were used to test the hypotheses at a 0.05 significance level. Results showed a significant multivariate effect of instructional method on combined ML competencies (Pillai’s Trace = 0.602, <i>p</i> &lt; 0.001). Follow-up univariate tests revealed that AI-based instruction produced significantly higher achievement in data modelling, data preprocessing, and data evaluation, with large effect sizes (η² = 0.341–0.461). However, no significant difference was found in data model construction. The findings demonstrate that AI-supported instruction substantially enhances students’ practical ML abilities more than video-based instruction. The study recommends integrating AI tools into ML curricula to strengthen practical competence.</p>