Loneliness among university students in Yogyakarta, Indonesia, is a growing mental health concern linked to psychological distress and suicidal ideation. Psychological First Aid (PFA) is a recognized approach for mitigating distress, yet access to human-led interventions remains limited. AI-powered chatbots offer a scalable alternative, but their effectiveness in loneliness reduction is underexplored. This study evaluates the effectiveness of an AI-driven WhatsApp chatbot in delivering PFA and reducing student loneliness. A mixed-methods randomized controlled trial was conducted among 32 university students in Yogyakarta (aged 18–22) with UCLA Loneliness Scale scores ≥35. Participants were randomly assigned to an experimental group (n = 16, PsyBot interaction) or control group (n = 16, neutral activity). The intervention involved a 30-min text-based PFA session via PsyBot. Pre- and post-intervention loneliness scores were analysed using paired and independent t-tests. Thematic analysis examined chatbot-user interactions. The experimental group showed a significant reduction in loneliness (M = 53.1 to M = 47.3; t (15) = 3.12, p = 0.007); no change was observed in the control group. Thematic analysis confirmed PsyBot's adherence to PFA principles by detecting distress (Look), providing empathetic responses (Listen), and offering coping suggestions (Link). AI-powered PFA shows promise in addressing student loneliness and increasing mental health accessibility. Future studies should evaluate long-term effects, cultural adaptability, and integration into institutional services. • PsyBot is a WhatsApp-based chatbot for digital Psychological First Aid. • The intervention targets student loneliness via brief AI-guided support. • PsyBot reflects Look, Listen, and Link principles from the PFA model. • Users reported emotional relief and felt more willing to seek help. • PsyBot offers scalable, low-intensity support for student mental health.
Stored only when both papers are in the corpus — 10 citations exist upstream.
Imported from OpenAlex (CC0) — nothing here is generated.