From Tutor to Confidant: Use Genres and Moral Economy in Student ChatGPT Domestication

How do students integrate conversational AI into academic life? Despite rapid adoption—teen ChatGPT use for schoolwork doubled between 2023 and 2024—empirical research examining actual usage patterns remains limited. This study addresses that gap by analyzing 10,536 conversational prompts from 36 undergraduate students across a year-long dataset of ChatGPT interactions. Drawing on domestication theory and Bakardjieva's use genres framework, we identify five distinct genres through which students transform general-purpose AI into specialized academic resources: academic workhorse, emotional companion, repair and negotiation, metacognitive partner, and trust calibration. Students exercise sophisticated agency in constructing fluid AI relationships, positioning ChatGPT variously as tutor, confidant, debugging partner, and system requiring conversational repair. Our findings reveal that trust depends not merely on accuracy but on communicative accountability—how ChatGPT fails matters as much as whether it fails. When repair rituals succeed, students return to collaborative interaction; when failures persist, they shift to adversarial verification testing. We argue for reconceptualizing AI literacy beyond technical fluency to encompass relational AI literacy: the capacity to manage the emotional, epistemic, and social dynamics of human-AI collaboration. Students' genre portfolio management, strategically selecting interaction patterns based on contextual needs, represents advanced domestication competence extending far beyond prompt optimization. Recognition of the invisible algorithmic labor students perform to make AI systems effective challenges instrumental framings of human-AI interaction and has implications for educational design and policy. These findings contribute to understanding how students appropriate emerging technologies through creative bricolage within the moral economy of academic life.