Search engines enable merchants to associate on-line ads with relevant keywords (for instance, through [Adwords].) Both merchants and search engines benefit economically from the merchants having as many appropriate keywords as possible. The space of keywords is large - it is useful to algorithmically generate keywords for merchants. Information about keywords is scarce, but it is known which merchants are bidding on which keywords. This paper describes WordRank, a method for using the bid landscape to infer a metric for keyword/merchant relevance, loosely based on the PageRank[Brin 98] algorithm.