Supportive or Persuasive? : A qualatative study on how Swedish university students interpret AI recommendations

Background: As AI recommendation systems become increasingly embedded in digital platforms, a central question arises: Do users perceive these systems as helpful assistants or as persuasive tools designed to influence behavior? While previous research has emphasized technical performance, less attention has been paid to how consumers interpret the intent behind AI recommendations. Purpose: This study aims to describe how Swedish university students interpret AI recommendations. Method: The research follows a qualitative, inductive approach, using semi-structured interviews with Swedish university students as the primary method for data collection. A total of three core interviews were conducted, followed by two additional validation interviews. Data were analyzed through thematic analysis, grounded in existing theoretical models such as the Technology Acceptance Model (TAM), Elaboration Likelihood Model (ELM) and Persuasion Knowledge Model (PKM). Findings: Participants consistently distinguished between AI recommendations perceived as genuinely helpful and those seen as manipulative or commercially driven. Supportive recommendations were associated with logical relevance, transparency and contextual alignment, while persuasive ones often involved emotional cues or repeated targeting. Some recommendations were experienced as both helpful and influential, indicating a complex interplay between assistance and persuasion. Participants also demonstrated strategic adaptation by intentionally interacting with the system to shape future outputs. Conclusion: The study reveals that students actively interpret and negotiate the intent behind AI recommendations, rather than passively accepting them. Trust and acceptance are highly context-dependent and influenced by perceptions of transparency, relevance and autonomy. These findings underscore the need for AI systems to be designed with ethical consideration, user control and contextual sensitivity in mind. Keywords: AI recommendations, user perception, persuasion, supportiveness, technology acceptance, qualitative study, trust, digital ethics, Swedish university student perspective