JISARA

Abstract: Understanding factors that influence user trust in Large Language Models (LLMs) is critical for successful AI adoption and appropriate use. This exploratory study provides the first empirical validation of the Acceptance-Trust Model (ATM) Framework proposed by Money and Thanetsunthorn (2025) regarding trust determinants in ChatGPT-like systems. We conducted a cross-sectional survey study with 94 participants examining relationships between self-efficacy, perceived control, perceived usefulness, perceived ease of use, LLM usage familiarity, and trust in LLMs. Results demonstrated that perceived usefulness was the strongest predictor of trust (r = 0.515, p < 0.001), followed by perceived ease of use (r = 0.438, p < 0.001), with four of five hypotheses receiving empirical support. The moderate trust levels observed (M = 2.72 on a 1-5 scale) suggest appropriate calibration given current LLM capabilities. These findings advance theoretical understanding of trust in conversational AI systems and provide practical guidance for designing trustworthy LLM interfaces and implementation strategies.

Download this article: JISARA - V19 N4 Page 38.pdf

Recommended Citation: Money, W.H., Mew, L., (2026). Trust in Large Language Models: An Exploratory Validation Framework. Journal of Information Systems Applied Research and Analytics 19(4) pp 38-52. https://doi.org/10.62273/NPBZ2264