Recent work on racial residential segregation shows how individuals’ perceptions of neighborhoods influence their housing choices and therefore contribute to the reproduction of segregation. Rental housing advertisements are a form of public discourse about neighborhoods which both influence and reflect those perceptions. I use n-gram regression and Structural Topic Models (STM) to investigate whether and how rental listings from the Seattle metropolitan area Craigslist page differ in association with neighborhood racial proportion. Neighborhoods with higher White proportion are associated with inviting words like ‘restaurant’ and ‘charming,’ while less-White neighborhoods are connected to security terms like ‘gated’ or transportation terms like ‘light rail’. STM and qualitative analysis shows that listings from White neighborhoods emphasize connections to neighborhood history and culture, while listings from non-white neighborhoods offer more incentives and focus on development features, sundering these areas from their surroundings. Finally, analysis of security discourse reveals that not only is language about security more common in less White neighborhoods, but administrative data show that actual security systems are less common. Without mentioning race, these listings reveal racialized neighborhood perceptions which likely impact individual’s neighborhood choice in ways that contribute to housing segregation.