AI chatbots in mental Health: How emojis, prompt type, and interactivity shape user perceptions in the United States and China

Abstract

Artificial intelligence (AI) chatbots are increasingly used in mental health support, but it remains unclear how key communication factors, such as emoji use, prompt type, and interactivity, shape user perceptions of chatbot messages and whether these effects differ by country. This study conducted a 2 (emoji use: present vs. absent) × 2 (prompt type: self-disclosure vs. direct asking) × 2 (interactivity: single-turn vs. multi-turn dialogues) between-subjects online experiment in the United States (N = 394) and China (N = 401). Participants evaluated ChatGPT's responses in a simulated mental health support scenario. Results show that U.S. participants responded negatively to emoji-present messages: Compared to chatbot messages without emojis, those containing emojis were rated lower in information quality, and when paired with self-disclosure prompts, led to reduced behavioral intention to use AI chatbots in mental health contexts. In contrast, interactivity emerged as a key driver of positive perceptions among U.S. participants. Multi-turn dialogues improved U.S. participants' evaluations of information quality, perceived support, and behavioral intention than single-turn dialogues. Chinese participants' evaluations, however, remained stable across all measures regardless of emoji use, prompt type, or interactivity. Chinese participants consistently reported more favorable perceptions and stronger behavioral intentions toward the AI chatbots than U.S. participants. These findings shed light on the nuanced roles of communication factors in shaping user perceptions and acceptance of AI-mediated mental health support across countries.