<p>Instructional videos have been widely implemented in educational settings. This study used “chemical bonds” as learning material to test the effects of different concept map strategies and cognitive prompts on the learning performance of middle school students through instructional videos. Employing a quasi-experimental design, Experiment 1 (<i>N</i> = 99) examined the impact of generating different concept maps (the Novak concept maps vs. the multidimensional concept maps) during instructional video learning. Results showed that generating multidimensional concept maps could enhance the quality of concept maps and contributed to higher transfer scores compared to Novak concept maps, but simultaneously resulted in lower retention scores. Experiment 2 (<i>N</i> = 197) employed a 2 (concept maps: the Novak concept maps vs. the multidimensional concept maps) × 2 (cognitive prompts: with vs. without) between-subjects quasi-experimental design to test the effects of cognitive prompts when learners generated different types of concept maps. Results showed that multidimensional concept map groups scored higher on concept map quality compared to Novak concept map groups. Furthermore, Novak concept map groups demonstrated better retention than multidimensional concept map groups. Moreover, participants who generated Novak concept maps with cognitive prompts scored higher on the transfer test than those who did not receive cognitive prompts. However, providing cognitive prompts hindered learners’ transfer performance when generating multidimensional concept maps. The findings of both experiments provided empirical support for the design of concept maps in instructional video learning.</p>