The improvising machine, Voyager, was composed by pioneering theorist and improvisor George Lewis in the 1980s. Sitting at the nexus of action and wider events, improvisation is a fertile field in which to conduct the Turing test. Here, we see whether an AI system can convince someone it is intelligent by responding to a complex environment. Although spontaneously produced improvisors rely on their trained behaviour to respond to unforeseen contributions. Drawing a parallel between the programming of an improvising machine and the habits of a human improvisor, Lewis’ denotation of Voyager as a composition seemingly threatens improvisation’s aptitude for self-expression and creativity. Through the example of Voyager, this article examines the relationship between habit and agential responsibility. I argue that the novelty of improvisation lies in the improvisor learning about new patterns in the musical material, or about themselves, as their habits are extended by unpredictability.