Measuring TAM from Online Customer Reviews: An LLM-based approach across IoT product categories with different levels of perceived autonomy

Online customer reviews offer accounts of consumers’ experiences with smart home technologies, but extracting theoretically defined aspects of technology acceptance from unstructured text remains challenging. This study examines the use of a zero-shot large language model (LLM) protocol to derive Technology Acceptance Model (TAM) related indicators from reviews. The corpus comprises 282 Amazon UK reviews across three product categories selected to represent different forms of device autonomy: robotic vacuum cleaners, smart speakers, and connected kitchen appliances. Using explicit construct definitions and seven-point behavioural anchors, the protocol assigns review-level scores for perceived usefulness, perceived ease of use, attitude toward using, and behavioural intention to use. The analytical strategy examines relationships among these indicators and compares their profiles across product categories. The study explores the potential of theory-guided LLM scoring for analysing post-purchase evaluations, while recognising that relationships among scores do not independently establish construct validity and that category comparisons cannot isolate the effect of autonomy.