Large language models in an app: Conducting a qualitative synthetic data analysis of how Snapchat's 'My AI' responds to questions about sexual consent, sexual refusals, sexual assault, and sexting

Key Findings

• — My AI's responses to sexual consent, refusal, and assault questions were generally consistent with sexual health education literature and pointed users toward trusted adults or resources
• — Responses were often succinct and somewhat generalized, with variability in reading level, tone, and depth across similar question types
• — The authors argue chatbot responses to sexual-assault-adjacent disclosure did not consistently reflect trauma-informed communication practices

Methodology Notes

Qualitative synthetic data analysis: 15 researchers independently submitted an identical set of standardized questions to Snapchat's My AI; outputs were coded via qualitative content analysis. A subset of questions was also run against Meta's Llama and Google's Gemini for comparison. Self-report/synthetic-query design; not a study of real user disclosures.

Sources