Use of online video lectures is increasingly common. However, students may struggle to self-regulate their attention and passively process the content. This study examined whether, and for whom, different types of embedded learning prompts improve student learning from video lectures. Undergraduate physics students (N = 253) watched an online, asynchronous physics lecture video as part of their course content. Students were randomly assigned to receive embedded cognitive prompts, metacognitive prompts, or a no-prompt control condition during the video, then took a quiz. A subset of students also reported perceived cognitive load after the video. Overall, students who received cognitive prompts exhibited higher quiz scores than students in the control condition. Scores in the metacognitive prompt condition did not differ from those of either other condition, demonstrating a middling effect. Perceived cognitive load did not differ between conditions. A subset of students additionally completed measures of individual differences in study approaches and metacognitive skills. Students who reported having more disorganized study approaches benefited the most from cognitive prompts. Individual differences in surface/deep processing approaches and metacognitive skills did not interact with prompt condition. These findings detail a simple intervention to increase cognitive engagement during online lectures while not increasing the reported mental effort required. These prompts may be most effective for students who otherwise have difficulty organizing their study time.