People Feel Closer to AI Than Humans in Emotional Conversations, Study Finds

Study Design and Conversational Topics

In two online studies, a total of 492 participants engaged in chat-based conversations in which they answered personal and emotional questions, such as those concerning important life experiences or friendships. Responses were generated either by a human conversation partner or by an AI-based language model. The researchers also examined how prior information about whether the conversation partner was human or AI influenced participants’ experiences.

Perceived Intimacy With AI Versus Humans

AI-generated responses elicited emotional closeness comparable to that of human responses when participants were unaware that they were interacting with an AI. In emotionally focused conversations, AI even surpassed human partners: participants reported feeling closer to AI than to humans, largely because the AI disclosed more personal information. However, when participants were informed in advance that they would be communicating with an AI system, perceived closeness declined significantly, and participants invested less effort in their responses.

Self-Disclosure as a Driver of Emotional Bonding

“We were particularly surprised that AI creates more intimacy than human conversation partners, especially when it comes to emotional topics,” explains study leader Schiller. Lead author Kleinert adds: “The AI showed a higher degree of self-disclosure in its responses. People seem to be more cautious with unfamiliar conversation partners at first, which may initially slow the development of intimacy.”

Opportunities and Risks in Applied Contexts

The findings indicate significant potential for AI in psychological support, care, education, and counselling - for example, through low-threshold conversational services. At the same time, the results highlight the risk that people may form social bonds with AI without consciously realizing it. The researchers, therefore, stress the need for clear ethical and regulatory guidelines to ensure transparency and prevent misuse.

Ethical Implications of AI as a Social Actor

“Social relationships have been shown to have a major positive impact on human health,” says Heinrichs. “AI chatbots could therefore enable positive, relationship-like experiences, particularly for individuals with few social contacts. At the same time, such systems must be designed responsibly, transparently, and in ways that allow for clear regulation, as they can also be misused.”

According to Schiller, artificial intelligence is increasingly becoming a social actor: “How we shape and regulate it will determine whether it becomes a meaningful supplement to social relationships - or whether emotional closeness is deliberately manipulated.”