Omnichannel Integration, Trust, and Purchase Intention: Evidence from Silicon, Hybrid Digital-Twin, and Human Experiments
(A2026-133770)

AUTHORS

Walter Assis, Universidade Federal do Espírito Santo - UFES; MARCELO BRANDAO, Federal University of Esp�rito Santo; Lelis Espartel, IADE - Universidade Europeia/CETRAD - Centre for Transdisciplinary Development Studies; Arthur Sarcinelli, Lus�fona University; Ricardo Limongi, Federal University of Goias, Brazil

ABSTRACT

This paper examines how perceived omnichannel integration shapes purchase intention through trust and tests whether AI-based silicon and digital-twin samples can reproduce this mechanism. A 2×1 fashion retail scenario experiment was implemented across three data-generating processes: a silicon sample from a large language model, a digital-twin sample based on detailed human profiles, and a human dataset. Across all datasets, higher integration increased trust, which in turn raised purchase intention, supporting a consistent mediation mechanism. Standardized effects were similar, and cross-sample Z-tests showed no significant differences. However, the digital-twin sample suggested full mediation, whereas the silicon and human samples retained a direct effect. These findings reinforce trust as a central pathway in omnichannel retailing and indicate that carefully designed AI-assisted samples can approximate human patterns in theory testing.