Context Shapes Control: A Survey of Generative AI Interaction Challenges Across Professional Groups

Abstract. Generative AI tools are now embedded in everyday professional practice, yet their interaction design assumes a uniform user. We surveyed 331 active users across six professional groups: academics, students, tech professionals, designers, non-technical workers, and artists. Combining reflexive thematic analysis (Braun & Clarke, 2006; 2021) of 395 open-ended responses with quantitative group comparisons, we mapped the challenges users encounter and the changes they want. Our thematic analysis identified shared challenges across general AI tools (output quality failures, interaction friction, limited control, and opacity) and image-based AI (visual inaccuracies, prompt translation difficulty, and insufficient iterative control). When asked what they would change in AI tools, participants across all groups prioritized control: they want mechanisms to guide AI behavior, verify sources, manage data, and make targeted adjustments without full regeneration. Quantitative analysis contextualized group differences in how they use and engage with Generative AI tools, including AI familiarity, tool usage, frequency, and how control needs are expressed. This confirms the need for an upcoming experiment testing whether context-adapted AI interfaces can address these control gaps and identifies which specific control mechanisms to prioritize in that design.