Do You Believe the Machine? Human Trust and Perception of LLM-Generated Content Across Twelve African Countries

This study examines how people across twelve African countries perceive and trust content generated by Large Language Models (LLMs) compared to human-authored content. It is, to our knowledge, the first large-scale controlled empirical study of LLM trust perception conducted with African populations. Almost all prior research on this topic has been conducted with Western, Educated, Industrialised, Rich, and Democratic (WEIRD) populations, creating a significant gap in understanding how trust dynamics operate in African contexts characterised by mobile-first internet access, diverse language communities, and varying urban/rural connectivity profiles. We employ a pre-registered mixed experimental design in which participants evaluate African-contextualised text passages across three content domains (news, science, legal) under two disclosure conditions (disclosed authorship vs. blind). We measure trust across three subscales (competence, integrity, benevolence), perceived credibility, information quality, and fact-checking intention. We additionally capture African-specific contextual variables internet access type, urban/rural classification, primary language, and news consumption platform as individual-difference moderators.