How Transparency Labels & Accuracy Affect Trust & Engagement on Social Media Content

This master's thesis examines how transparency labels and factual accuracy influence user trust and engagement with social media content. It uses a three-study mixed-method design: an observational analysis of 200 Reddit posts, a survey experiment with 341 valid respondents, and 15 semi-structured interviews. The findings show that users do not react to AI involvement in a simple AI-versus-human way. Human-created content received the strongest trust-related and behavioural evaluations, fully AI-generated content received the weakest evaluations, and AI-assisted content occupied a more moderate position close to the no-label condition. The thesis also shows that factual accuracy, content context, media format, and perceived human oversight strongly shape how users judge and engage with labelled content. Overall, the thesis argues that transparency labels should be more precise and should explain the degree and form of AI involvement rather than only stating that AI was used.