This Article addresses 'spamdexing,' namely, the practice of stuffing invisible keywords into webpages in order to try to get more favorable listings with search engines. For instance, some website owners will stuff the trademarks of competitors into a webpage’s code, particularly by using 'meta tags,' indexing keywords that can be hidden in a webpage’s source code. Although meta tags are not typically viewed by users, the code can be read by search engines, with the result that webpages may be improperly boosted in search engine rankings. Such practices can confuse the public and have also spurred trademark lawsuits. But the real harm of spamdexing is best explained not by law, but instead by information science. Namely, spamdexing introduces 'noise' into search engine databases, rendering searches 'imprecise' - i.e., overinclusive. With that in mind, the Article proposes looking to relevance and reasonable expectations. Specifically, it is necessary to consider the relevance of a website user’s search terms to the search goal, and of a website owner’s meta tags to actual website content. It is also necessary to consider the reasonableness of each party’s desire to use such terms for searching and indexing purposes, along with the interests of trademark owners. There is a tension between these interests, as searchers and website owners will tend to externalize the costs of information dissemination and retrieval, and trademark owners may sometimes object to legitimate uses of meta tags. Finally, the Article provides a broader view of new forms of dissemination and retrieval, examining the roles that libraries have traditionally played, and those that the Internet-as-library may yet play.