<![CDATA[Occupational health and safety (OHS) education in vocational high schools requires higher-order thinking skills (HOTS), yet formative assessment is often limited by delayed teacher feedback. This study examined whether generative AI-supported formative assessment using ChatGPT and Gemini improved HOTS and learning motivation compared with conventional teacher-led assessment, and explored the relationship between perceptions of AI-assisted assessment, motivation, and HOTS. A posttest-only nonequivalent control group quasi-experimental design involved 60 Grade X Machining Engineering students at SMK Negeri 1 Padang, divided into two intact classes of 30 students each. HOTS was measured using a 20-item reasoned multiple-choice test covering Bloom’s cognitive levels C4–C6, while motivation and perceptions of AI-assisted assessment were measured using validated questionnaires. Independent-samples t-tests were used for the primary analysis, followed by exploratory partial least squares structural equation modelling (PLS-SEM) in the experimental group. No significant differences were found between the AI-assisted and conventional assessment groups in HOTS scores (81.73 vs. 80.87, p = .417, d = 0.21) or motivation (54.03 vs. 54.83, p = .497, d = −0.18). Within the experimental group, perceptions of AI-assisted assessment were positively associated with motivation (β = 0.743, p < .001), and motivation was positively associated with HOTS (β = 0.476, p = .023). Although the indirect effect was statistically significant (β = 0.354, p = .042), its bias-corrected confidence interval included zero, making the mediation evidence inconclusive. These findings do not demonstrate that AI-assisted formative assessment improves HOTS or motivation beyond conventional teacher feedback. However, the observed motivational relationships warrant further investigation using larger samples, baseline measurements, and longer interventions.]]>
