How emotional response styles of generative AI companions shape adaptive emotion regulation and short-term AI reliance

Key Findings

• — Both response styles reduced immediate negative affect to the same degree (b = 0.055, p = 0.323).
• — Adaptive regulation-oriented replies produced higher emotion-regulation self-efficacy (b = 0.500, p below 0.001), better independently coded strategy quality (b = 0.353, p = 0.009) and greater affective recovery without AI (b = 0.279, p below 0.001).
• — Relational reassurance-oriented replies produced greater perceived AI emotional necessity (b = 1.010) and stronger short-term AI-reliance intentions (b = 0.628, both p below 0.001).
• — Bootstrap models were consistent with an adaptive pathway through self-efficacy (indirect effect 0.279, 95% CI 0.171 to 0.402) and a reliance pathway through perceived necessity (0.375, 95% CI 0.264 to 0.501); the author calls the pattern 'immediate relief, subsequent divergence'.
• — Reliance outcomes are framed as short-term state measures, not clinical dependence.

Methodology Notes

Randomized experiment, n = 380 adults recruited through Wenjuanxing (Questionnaire Star), campus mailing lists and notices at Guangzhou College of Commerce; companion built on OpenAI GPT-5.5 (fixed snapshot gpt-5.5-2026-04-23, low reasoning, 300-token replies) with matched interface; mediation pathways are not randomized; single author affiliated with the School of Art and Design, Guangzhou College of Commerce; conducted in Chinese. Published 8 September 2026 (online date 24 August 2026), CC BY.

Sources