Videos are an engaging medium for learning as they provide affordances beyond text or audio-only, thereby allowing the creator more flexibility for content generation. Easy access to videos on the Web and their popularity within everyday discourse has made them an accepted medium for teaching and learning. However, this increase in the availability of videos makes it challenging for an instructor, teacher, or learner to assess their viability for learning. Each search can lead to thousands of videos on a given topic. Although content platforms and search engines use a range of data and algorithms for the recommendation, these are not tailored to recommend video content specifically for learning. In this paper, we present and test a theoretically motivated model to rate videos on their potential to support learning goals. Visual Appeal, Understanding of Content, Engagement with Topic, and Recommendation Preference (VUER) are the four components of our proposed model. We assess the model on two content areas, "Big Data and Augmented Reality," using videos from the TED (Technology, Entertainment, and Design) platform. Using crowdsourcing, we collect ratings for items in the VUER model and assess effectiveness through the correlation between 1) model elements and self-reported learning and 2) the rank of a video in the TED search and self-reported learning. Results show that the VUER model ratings of videos strongly correlate with the expected learning gain of the users for videos on both topics. The expected learner ranking and observed TED ranking of videos do not necessarily align, suggesting that the ranking on platforms such as TED is not a direct fit to judge the teachability of the video content. Thus, there is a need to create personalized systems to support online video curation tasks for learners and teachers.