A corpus-based analysis of politeness strategies in Arabic speakers’ interactions with ChatGPT

This study investigates how Arabic speakers employ politeness strategies when interacting with the conversational AI system ChatGPT, focusing on speakers’ own politeness behavior in human–AI communication. Although politeness in AI interaction has received increasing scholarly attention, prior research has largely concentrated on AI-generated politeness, leaving the politeness strategies used by human speakers toward AI largely unexplored, particularly in Arabic-speaking contexts. The study addresses three questions concerning the types of politeness strategies Arabic speakers use in prompts, their distribution across discourse segments (openings, bodies, and closings), and the influence of interactional features such as topic shifts. Data were drawn from the WildChat dataset, from which 349 Arabic-dominant conversations containing a minimum of ten turns were extracted. User turns were coded and analyzed using an adapted Brown and Levinson framework combining quantitative frequency analysis with qualitative pragmatic interpretation. The findings reveal a strong preference for bald on-record strategies, accompanied by limited but systematic use of conventional indirectness and relational markers. The pattern is consistent with an interactional reorientation away from elaborate face redress and toward task-oriented efficiency in AI-mediated communication.