This article reports on two studies designed to examine the landscape of online streamed videos, and the features that may support vocabulary learning for low-income preschoolers. In Study 1, we report on a content analysis of 100 top language- and literacy-focused educational media programs streamed from five streaming platforms. Randomly selecting two episodes from each program, we identified the prevalence of vocabulary opportunities, and the pedagogical supports--techniques or features in these media that are designed to orient children to specific vocabulary words. In over the 2,000 scenes coded, we identified two overriding categories of supports: ostensive cues, designed to provide definitional information to children; and attention-directing cues, designed to signal children's attention to a target word. In Study 2, we use eye-tracking technology to examine which of these pedagogical supports might predict children's ability to identify program-specific vocabulary. Results indicated that although ostensive cues predicted overall attention to scenes, attention-directing cues were most effective in directing children to target words and their subsequent word identification. Children with higher language scores were more likely to use these cues to their advantage than their lower language peers. These results may have important implications for designing digital media to enhance children's opportunity to learn vocabulary. [This article was published in "Journal of Educational Psychology" (EJ1201278).]