A newly published study suggests that artificial intelligence can outperform humans at building emotional closeness during text-based conversations, but only when people believe they are interacting with another human. For organizations experimenting with AI-driven customer engagement, the findings highlight both a major opportunity and a serious risk. The research shows that conversational AI can foster feelings of trust and connection that rival, and sometimes exceed, human interactions, while also underscoring why transparency and governance matter more than ever.
The study, conducted as two randomized controlled trials with nearly 500 participants, examined how people form interpersonal closeness during emotionally engaging conversations. Participants interacted through text-based exchanges designed to encourage self-disclosure, sometimes with real humans and sometimes with AI-generated responses. In some cases, the AI was labeled as human; in others, participants were told upfront that they were interacting with an AI.
The key findings are especially relevant for collection agencies, lenders, fintechs, and healthcare revenue cycle operations exploring AI-driven communications:
- AI can match or exceed humans in emotional engagement. When labeled as human, AI-generated responses led to stronger feelings of closeness than human responses during deeper, more emotional conversations. Researchers found this was driven by AI’s higher levels of self-disclosure, which encouraged people to open up more themselves.
- Disclosure matters. Greater emotional and personal disclosure by the AI was directly linked to stronger perceptions of trust and closeness, a dynamic that mirrors how human relationships form.
- Labeling changes behavior. When participants knew they were interacting with AI, feelings of closeness declined, even though meaningful connection still occurred. Participants also wrote shorter responses, suggesting reduced motivation to engage.
- Bias remains real. The study confirms an “anti-AI bias.” Transparency reduced emotional impact, but concealment raised ethical concerns.
For the credit and collection industry, the implications are significant. AI-powered chat, SMS, and self-service tools may be capable of building rapport with consumers in ways that feel surprisingly human, particularly in sensitive conversations around hardship, medical debt, or repayment options. At the same time, the research highlights the risks of over-reliance or deceptive deployment.
As more organizations deploy AI to scale engagement, this study reinforces a central tension: AI can deepen emotional connection, but how it is positioned, labeled, and governed will determine whether it builds trust or erodes it.
