It is well investigated that the expression of racial prejudice is often induced by news coverage on the internet, and the exposure to media contributes to the cultivation of long-term prejudice. However, there is a lack of information regarding the immediate effects of news delivered through television or television-like media on the expression of racial prejudice. This study provides a framework for understanding such effects by focusing on content-audience associations using the logs of an “online television” service, which provides television-like content and user experiences. With these logs, we found an association between the news-watching and comment-posting behaviors. Consequently, logs relevant to two distinct forms of racism, modern and old-fashioned racism, were extracted. Using mathematical modeling, which considers the different levels of program inducements to racist expression, personal inclinations of audiences to racism, and certainty of prediction of audience behaviors, we found three associative patterns between the news programs and audiences. The relevance of the topics covered to the basic beliefs of each form of racism was characterized into three clusters: expression as a reaction to news that is directly relevant to the basic beliefs of racism with weak inducements by non-bigots, minority abuse by distorting the meanings of news content indirectly relevant to the beliefs but with strong inducements by audiences with a strong bias, and racial toxic opinions independent of the news content by clear bigots. Our findings provide implications for inhibiting the expression of online prejudice based on the characteristics of these patterns.