Document Type


Publication Date



The likelihood of confusion standard defines the scope of trademark infringement. Likelihood of confusion examines whether there is a substantial risk that consumers will be confused as to the source, identity, sponsorship, or origin of the defendants’ goods or services. This Article presents a contemporary empirical analysis of the various factors and how they interact. Conventional wisdom teaches us that courts should comprehensively traverse each factor and that likelihood of confusion cases generally require jury determination. However, the data reveals that neither is true. Instead, courts provide early off-ramps to litigants by “economizing,” and analyzing only a handful of factors or by “folding” factors within each other. The findings also reveal (1) which forums are pro-defendant and which are pro-plaintiff; (2) the impact of rivalry and fair use on outcomes; and (3) an apparent Ninth Circuit dominance.

What constitutes “confusion” remains highly subjective and difficult to evaluate. Proxies like intent, survey evidence, mark strength, and consumer sophistication fail to incorporate real-world purchasing conditions or are better considered within omnibus factors. In contrast, actual confusion, mark similarity, and competitive proximity provide judges with a potent trio of factors to guide the infringement inquiry. Together with safe harbors for descriptive and expressive uses, these rules of thumb enable courts to resolve trademark disputes more coherently, consistently, and expeditiously. This Article concludes with a blueprint of how these rules of thumb complement artificial intelligence systems and how those systems can use empirical studies as training data to inform future likelihood of confusion analyses.

Publication Title

American University Law Review