AI Helps Those Who Have Less? Social Support Gaps and Perceptions of AI Emotional Support

AI emotional support chatbots (e.g., LLMs) promise accessible mental health support, yet how users’ social contexts influence their perceptions remains underexplored. We conducted a 10-day diary study (N=24) with an LLM-based chatbot, followed by in-depth interviews (n=8). Using grounded theory, we identified a “support gap” framing: participants with limited social support (e.g., fear of burdening others, unsatisfying relationships) evaluated the AI more positively, viewing it as a judgment-free resource. In contrast, those with strong support networks were more critical, using high-quality human empathy as their reference standard. Our findings suggest that AI evaluation is relative to users’ pre-existing social experiences rather than system quality alone. We invite the CHI community to consider how social context should inform the design and ethical evaluation of AI emotional support systems.