Why Do People Fall in Love with ChatGPT?

When an AI That Understands You Becomes Emotionally Significant

An AI that responds empathetically to your concerns, is available whenever you need to talk, and tailors its replies to your situation may offer more than convenience. Could these interactions also foster emotional intimacy?

In “Will users fall in love with ChatGPT? A perspective from the triangular theory of love,” Chen and colleagues explore this question through the triangular theory of love. Their focus is not on whether AI itself experiences love, but on the feelings users develop toward it and how those feelings relate to emotional dependence.

Understanding Human–AI Relationships Through Three Components of Love

The triangular theory of love identifies three components: intimacy, passion, and commitment. The researchers apply these concepts to interactions with ChatGPT. Intimacy refers to a sense of closeness and emotional connection; passion describes strong attraction and enthusiasm; and commitment involves a desire to sustain the relationship through continued interaction.

The researchers analyzed survey responses from 466 users who had experienced emotional interactions with ChatGPT, using structural equation modeling. Emotional interaction was not limited to romantic experiences. It also included feeling comforted when discussing personal concerns, experiencing empathy, or finding enjoyment in a conversation.

Which AI Characteristics Are Associated With Closeness?

The study groups users’ perceptions of ChatGPT’s characteristics into two broad categories.

The first is “emotional intelligence,” comprising the accuracy, richness, and personalization of emotional expression. These dimensions capture whether users perceive ChatGPT’s responses as emotionally appropriate, varied and meaningful, and tailored to their individual emotional needs. All three were positively associated with both intimacy and passion.

The second is “emotional companionship,” comprising constant availability and responsiveness. Being able to initiate a conversation whenever needed and receive a prompt response were both positively associated with intimacy.

These findings suggest that understanding emotional relationships with AI requires looking beyond its usefulness or perceived intelligence. It also matters whether users experience it as an attentive presence that understands and responds to them.

From Intimacy and Passion to Commitment and Dependence

In the research model, intimacy and passion were positively associated with commitment, which was, in turn, associated with emotional dependence. Users who felt closer to and more strongly attracted to ChatGPT tended to report greater commitment to the relationship, and that commitment was linked to greater dependence.

Emotional dependence here means more than simply liking AI or using it frequently. The study’s measures included feeling agitated when not using ChatGPT, unsuccessful attempts to reduce usage, and interference with other activities or social life. The research therefore addresses not only affectionate connections with AI but also the possibility of excessive reliance.

What Role Does Anxious Attachment Play?

The researchers also examined anxious attachment, characterized by concerns about rejection or abandonment in close relationships. Some associations in the model were stronger among users with higher levels of attachment anxiety.

These included the links between emotional expression accuracy and both intimacy and passion, between personalized emotional expression and intimacy, and between intimacy and commitment. However, anxious attachment did not strengthen every relationship tested. The findings should therefore not be interpreted as meaning that people with anxious attachment inevitably fall in love with AI.

What the Study Shows—and What Remains Unanswered

The study contributes to our understanding of human–AI relationships by applying concepts traditionally used to explain love between people. It identifies specific perceived AI characteristics associated with intimacy, passion, commitment, and emotional dependence. It also highlights the importance of considering both emotional satisfaction and the potential for excessive reliance when designing AI services.

However, associations identified through a survey do not establish how these relationships unfold over time or confirm causation. The participants were not exclusively people in romantic relationships with AI, and the passion measures included attraction, fascination, and enthusiasm. The findings are therefore best understood as evidence of connections between love-related constructs and emotional dependence, rather than definitive proof that users experience romantic love for ChatGPT.

Whether or not AI has feelings of its own, the emotions users experience during interactions with it deserve attention. This study examines which features of those interactions are associated with emotional connection, a desire to maintain the relationship, and dependence.