Study Overview

A recent randomized controlled trial has evaluated the effectiveness of the ChatGPT 4o-based therapeutic chatbot named Amanda for providing relationship support. The findings indicate that a single session with Amanda can be as effective as traditional journaling in resolving relationship conflicts. This study was conducted by Dr. Laura Vowels and her team from the University of Lausanne and the University of Roehampton, and published in the open-access journal PLOS Mental Health.

Key Findings

• The study involved 258 participants aged 18 and older, all in romantic relationships facing non-abusive conflicts.
• Participants were divided into two groups: one interacted with the chatbot Amanda, while the other completed a journaling task focused on conflict reappraisal.
• Both groups reported improvements in their specific relationship issues, overall relationship satisfaction, and personal well-being immediately after the intervention and two weeks later.
• No significant differences were found between the two groups, suggesting that Amanda’s chatbot therapy is comparable to journaling.

Study Limitations

The authors noted that the single-session format limited their ability to assess the chatbot’s potential for building a therapeutic alliance over time, which is often a key benefit of traditional therapy. Additionally, the exclusion of participants who did not return for follow-up may have skewed the results, potentially overestimating the effectiveness of both interventions.

Future Research Directions

The researchers advocate for further studies to explore the effectiveness of LLM chatbots like Amanda in multi-session formats and within clinical populations that require advanced risk detection capabilities.

Expert Insights

Dr. Vowels stated, “Our study indicates that a single session with Amanda can significantly enhance relationship satisfaction and individual well-being. This highlights the potential of large language model chatbots to provide accessible, evidence-based relationship support.” She also emphasized the positive feedback from participants regarding Amanda’s empathy and usability, suggesting that individuals are open to engaging with AI in sensitive contexts.

For more information, you can access the full study here.

Citation: Vowels LM, Vowels MJ, Sweeney SK, Hatch SG, Darwiche J (2025) The efficacy, feasibility, and technical outcomes of a GPT-4o-based chatbot Amanda for relationship support: A randomized controlled trial. PLOS Ment Health 2(9): e0000411. https://doi.org/10.1371/journal.pmen.0000411