Behind the hype: a closer look at voice assistant skills, product control strategy and user evaluations

Purpose Drawing on the assemblage framework and signaling theory, this research examines how a voice skill's product control strategy, specifically controllability variety and complexity, shapes perceived intelligence and user evaluations. It also explores the invocation strategy (licensing vs. co-branding) as a boundary condition for these relationships. Design/methodology/approach A multi-source, multi-method approach is adopted, using field data from 481 Amazon Alexa skills, over 47,000 user reviews, two user surveys and expert-coded data. Hypotheses are tested using multiple analyses in SPSS. Findings Controllability variety functions as a competence signal and is positively associated with perceived intelligence. Controllability complexity serves as a friction signal and penalizes perceived intelligence. In turn, perceived intelligence acts as a primary driver of user evaluations. Moreover, an invocation license intensifies the negative complexity–perceived intelligence relationship more than co-branding invocation does. Originality/value This study shifts the research focus from platform-level voice assistants to their micro-level skills. It integrates signaling theory with the assemblage framework, establishing that voice skills function as a critical entity within a triadic user–AI–product assemblage. By introducing and empirically contrasting the invocation strategy, it provides a novel perspective on how the locus of artificial intelligence identity governs attribution within the assemblage and shapes user perceptions of intelligence.