Decoding Consumer–AI Interaction Through Perceived Cognitive Capability and Anthropomorphism: Evidence From Unstructured Big Data

ABSTRACT This study investigates how consumers interpret and assess artificial intelligence voice assistants in everyday interactions. Using Task‐Technology Fit and Anthropomorphism Theory as interpretive lenses, it examines how perceived cognitive capability is evaluated in user‐generated content and how the alignment between artificial intelligence capabilities and user needs is reflected in consumer discussions. Using an unsupervised machine learning approach, we analyzed 150,020 Reddit posts from UK users discussing artificial intelligence voice assistants. The analysis identifies seven key topics that capture consumer evaluation, which are subsequently interpreted and labeled as aspects such as user experience, tech traits, intelligence, and social attraction. To examine evaluative variation, we conducted sentiment‐stratified topic co‐occurrence analysis, revealing a configurational pattern in which consumer evaluations emerge from combinations of interacting attributes rather than isolated factors. The findings show a semantic asymmetry in how identified topics related to artificial intelligence capabilities are interpreted across evaluative contexts: intelligence‐related attributes are associated with learning and responsiveness in favorable contexts but shift toward concerns about monitoring and surveillance in unfavorable contexts. Such context‐dependent patterns may remain hidden in studies that rely solely on predefined constructs. The outcomes suggest an interpretive mechanism whereby the same topic is evaluated differently depending on the evaluative context. This study contributes by developing a theory‐building explanation of consumer‐artificial intelligence interaction and by providing a foundation for future confirmatory research.