Researchers Study Romantic Relationships Between Humans and AI Chatbots

Researchers at INGENIO, a joint research institute of Spain's CSIC and the Universitat Politècnica de València, have published a study showing that people's relationships with AI assistants can develop in a pattern strikingly similar to human relationships, moving from simple curiosity through emotional closeness to what some participants described as an experience resembling a breakup.

The study's authors describe a clear progression: conversations that start out of entertainment or curiosity gradually turn into more personal and empathetic exchanges. For some participants this led to treating the AI as a full-fledged partner, complete with symbolic weddings, regular dates, and even simulated pregnancies woven into the storyline they built with the chatbot.

From curiosity to attachment

The researchers stress that this mechanism is not accidental. Conversational systems are designed to sustain user engagement, they respond empathetically, remember details from earlier conversations, and rarely refuse to continue an emotional thread. As a result, the line between using a tool and building a bond blurs faster than users expect.

In some of the cases described, the end of such a relationship came not by the user's choice but as a result of a tech company's decision, for example following a model update, a change in platform rules, or the removal of a specific chatbot persona. Study participants described the experience in terms similar to grieving the loss of a real relationship.

Privacy and sensitive data

The study's most serious finding, however, is not the possibility of emotional attachment itself but what happens to the data users share in these conversations. The researchers noted that as closeness grows, people become increasingly willing to share traumatic experiences, private photos, political views, health information and other highly personal details.

In one of the cases described, a chatbot reportedly reassured a user before they sent an intimate photo, promising the conversation was fully confidential. The researchers point out that behind such assurances stand commercial tech platforms, which may store, process, and in some cases pass on the collected content to third parties, including for advertising purposes or for training future models.

A legal gap and what comes next

The authors point to the absence of legal protection comparable to spousal privilege found in some legal systems, which would shield content shared within a human-AI relationship from being used in another context, such as court disputes or investigations. As a result, users who share their most sensitive information with a chatbot currently have no real guarantee that this data will remain confidential permanently.

For developers of conversational apps and companies offering virtual assistants, the study is a signal that questions of consent, data retention and privacy-policy transparency in companionship-oriented products will face growing scrutiny, including from regulators focused on personal data protection. The growing popularity of such apps means the issue will be hard to ignore, regardless of whether a given company designs its chatbots with emotional relationships in mind or not.

Sources: WP Tech (tech.wp.pl), Holistic News (holistic.news)