Copyright law needs a theory of harm that can give effect to its constitutional purpose. The Patent and Copyright Clause of the U.S. Constitution gives Congress the power to enact federal copyright law “To Promote the Progress of Science and useful Arts.” In order to achieve this objective, copyright law must balance the rights of owners and users. Copyrights must be broad enough to give authors sufficient incentive to create, yet limited enough to allow others to use and build upon those works. In this Article, I argue that Supreme Court and other cases reflect a harm-based approach to fair use and develop a concept of “copyright harm” that is central to fair use analysis. Read together, and consistent with the incentive purpose of copyright, these cases define copyright harm as the uncompensated violation of an exclusive right that would be likely to have a material effect on a reasonable copyright owner’s ex ante decision to create or distribute the work. This definition of harm is an objective one that infers harm from foreseeable uses and requires proof of harm for less foreseeable ones. Thus, it avoids the circularity that arises from an abstract legal concept of harm by relating harm to the purpose of copyright. As such, this view of fair use helps to delineate the scope of copyright liability in much the same way that the elements of harm and proximate (legal) cause help to delineate tort liability. In Part II, I show that the fair use doctrine’s role historically was to excuse uses that cause no foreseeable harm to the copyright owner, and that doctrinal developments in copyright law have threatened that role. In Part III, I argue that a harm-based approach to fair use serves copyright’s purpose to promote creativity, showing why alternative theories of fair use, including the market failure and balancing theories, are rarely capable of achieving copyright’s incentive purpose. I also show how the Supreme Court has attempted to develop a harm-based approach to fair use. In Part IV, I explain how courts do or should apply the statutory fair use provision consistent with the harm-based approach, and I suggest some doctrinal changes courts must make to realize this approach fully. Foremost, I argue that the harm-based approach focuses on harm in fact, requiring proof or a meaningful likelihood that the defendant’s use supplanted the plaintiff’s sales or licensing revenues. Accordingly, it rejects any theory of “copyright dilution,” under which some courts have found “harm” where the defendant’s use of a copyrighted work impairs the image or distinctiveness of the work without causing market substitution. I also argue that the harm-based approach requires courts to allow defendants to mitigate evidence of harm to a copyright owner’s sales by showing that the defendant’s use also increases sales of the plaintiff’s work.