Talking or typing? How GenAI modality, product type, and multitasking shape recommendation response behaviour
Mai Nguyen, Mojtaba(Moji) Barari, Yunen ZhangPurpose
This research aims to explore how generative AI (GenAI) recommendation modality (text vs voice) influences consumer recommendation adoption and how this effect further interacts with product type (search vs experience), processing fluency and multitasking to shape consumers’ behavioural responses to recommendations. Given that GenAI recommendation systems increasingly support interactive consumer decision-making, the authors examine modality as a key design feature shaping how GenAI-generated recommendations are delivered to consumers.
Design/methodology/approach
Across five studies, the authors examine how GenAI recommendation modality influences consumers’ recommendation adoption. The preliminary field study uses actual customer interactions in a natural commercial environment to establish the practical relevance of the phenomenon. Study 1 replicates the core modality effect in a controlled Chinese experiment. Study 2 examines the moderating role of product type. Studies 3 and 4 investigate processing fluency and multitasking as the underlying mechanism and boundary condition, using behavioural measures across all studies.
Findings
The findings demonstrate that GenAI recommendation modality affects consumers’ recommendation adoption, with text-based recommendations outperforming voice-based recommendations. The modality also interacts with product type and this effect is mediated by processing fluency: text-based recommendations elicit stronger behavioural responses for search products, whereas voice-based recommendations are more effective for experience products. Multitasking further moderates this pattern, favouring voice-based recommendations under high cognitive load across product types.
Research limitations/implications
The findings are based on controlled experimental settings and future work could examine additional contextual or individual-level moderators to enhance external validity.
Practical implications
Marketers should tailor GenAI recommendation modalities based on product type and anticipated multitasking environments. Text-based interactions are more suitable for detail-oriented search products, whereas voice-based interactions better suit experience products and multitasking contexts.
Originality/value
This research contributes to the GenAI literature by identifying recommendation modality as a critical yet understudied design feature of GenAI recommendation systems. The findings show that recommendation modality plays a central role in shaping consumer responses to GenAI-generated recommendations and extend processing fluency theory and task–technology fit by integrating product type and multitasking as key boundary conditions.