10 Questions to Fall in Love with A Chatbot: An Experimental Study on Interpersonal Closeness with Large Language Models (LLMs)

Authors

• Jessica M. Szczuka, INTITEC, Research Center Trustworthy Data Science and Security, Research Alliance Ruhr, University Duisburg-EssenFollow
• Lisa Mühl, INTITEC, University Duisburg-EssenFollow
• Paula Ebner, INTITEC, Research Center Trustworthy Data Science and Security, Research Alliance Ruhr, University Duisburg-EssenFollow
• Simon Dubé, Kinsey Institute, Indiana University, Bloomington, USAFollow

Abstract

Large language models (LLMs), like ChatGPT, are able to compute affectionately nuanced text, which can be deployed in online dating or power artificial companions. This study explores whether people can experience intimate interest and interpersonal closeness toward a potential human or AI partner, based on such text. In a mock matchmaking scenario, 307 participants rated 10 text responses to the Interpersonal Closeness Generation Task from the dating profile of their potential partner, unaware that all were LLM-generated. Surprisingly, the source label (human or AI) had no significant impact on intimate interest and interpersonal closeness. Instead, the perceived quality and human-likeness of responses shaped reactions. These findings advance research in human-machine communication, with social and ethical implications for the future of our AI-mediated intimacy.

DOI

10.30658/hmc.13.3

Author ORCID Identifier

Jessica M. Szczuka: 0000-0002-7722-8316

Lisa Mühl: 0009-0006-7434-8079

Paula Ebner: 0009-0001-3377-5027

Simon Dubé: 0000-0003-4947-7937

Recommended Citation

Szczuka, J. M., Mühl, L., Ebner, P., & Dubé, S. (2026). 10 questions to fall in love with a chatbot: An experimental study on interpersonal closeness with large language models (LLMs). Human-Machine Communication, 13. 75-100. https://doi.org/10.30658/hmc.13.3