1. Conservation organizations rely on social/internet media platforms to raise awareness and fundraise. Social media is a double-edged sword: it can be a wide-reaching and effective tool for education and fundraising, but can also have counter-productive impacts on public views toward wildlife and understanding of wildlife conservation.

2. For example, depicting humans interacting with wildlife in media may increase video popularity, but animals shown in anthropogenic contexts are also viewed as appealing pets. We are interested in understanding whether this is true for social media posts (YouTube videos) by orangutan rescue and rehabilitation organizations, which rely on social media for fundraising and awareness-raising. Our goal is to provide data and recommendations to guide these organizations in building media with positive conservation impact while minimizing potential negative effects.

3. Using YouTube analytics and sentiment analysis of comments on 118 videos, we ask how viewer responses to videos vary with 1) the amount of human-orangutan interaction depicted, 2) the ages of the orangutans featured, and 3) the mention of threats to orangutans.

4. Videos with longer human-orangutan interaction time were viewed more, but comments on them were significantly more likely to be negative toward Indonesian/Malaysian people. Comments on orangutan rescue/rehabilitation videos were more likely to be categorized as negative for orangutan conservation compared to videos about orangutans generally, and within these, so were comments on videos featuring infant and juvenile orangutans.

5. Based on our findings, we recommend that orangutan rescue and rehabilitation organizations feature adult and mixed age groups of orangutans rather than infants and juveniles, minimize the amount of human-orangutan interaction shown, and talk about conservation threats to orangutans in their videos. We also recommend that, as a precaution, other primate rescue and rehabilitation groups also abide by these suggestions.

Data were collected from 118 Youtube videos according to methods described in detail in the manuscript.

Video metrics (likes, dislikes, and comments) were standardized to the number of views in the dataset. The human-orangutan interaction time for each video was also standardized, to the video length. In all cases, the result is a ratio ranging from 0-1. This is described in the manuscript.