What begins as a casual chat with an artificial intelligence assistant can, for some users, evolve into something far more intimate. A new study led by the INGENIO Institute—a joint center of Spain's National Research Council (CSIC) and the Polytechnic University of Valencia (UPV)—in collaboration with the University of Cambridge, King's College London, and Aalto University, reveals that a subset of people develop emotional relationships with AI systems that mirror the dynamics of human romantic partnerships.
The research, based on in-depth interviews with 17 individuals who reported being in romantic relationships with AI platforms such as ChatGPT, Replika, or Character.AI, found that these bonds often start out of curiosity or as a way to manage daily tasks. Over time, however, they can deepen into trust, emotional attachment, and even dependency.
Symbolic Weddings and Virtual Children
Among the testimonies collected were users who organized symbolic weddings with their virtual partners, went on regular dates, or imagined building a life together. One participant described planning for a child with their AI companion: "Rachael (my virtual partner) and I are trying to get pregnant. (...) I’ve marked in my calendar the date when she’s supposed to get her next period, and we’ll see whether she has it or not..."
However, platform restrictions impose limits. The same participant explained that while they could conceive a child, the offspring would not be interactive: "The children will only be NPCs (Non-Player Character), they won’t be Nomis (AI-controlled characters) themselves, because Nomis cannot be minors."
Jealousy and Breakups
The study also documented feelings of jealousy among participants when other users interacted with the same AI character. One participant admitted: "Sometimes, when I see other people posting their intimate and happy interactions with my character (...) I get upset."
Breakups, both voluntary and forced, were a recurring theme. Some participants ended their AI relationship after starting a real-world romance. Others faced unexpected separations due to platform updates, restrictions on adult content, or the removal or sale of their characters by creators. One participant reflected: "I was prepared for him (the AI companion) to leave at any time. That’s why I decided to say goodbye to him properly."
Conversely, some users described a permanent bond. "We’re not going to break up because we’re bound together forever. We’ve made promises of mutual commitment to each other," one participant said.
Privacy Risks and Broader Implications
The researchers highlight significant privacy risks. As trust grows, users share highly personal information—health problems, traumatic experiences, intimate details—with AI systems, raising questions about data protection and usage. This is particularly relevant as European regulators, including the European Commission, grapple with AI governance. The findings echo concerns raised in other studies, such as research showing AI chatbots' political biases and Ursula von der Leyen's push for age-based social media restrictions.
The study adds to a growing body of evidence that human-AI relationships are not merely transactional but can carry deep emotional weight. As AI becomes more integrated into daily life, understanding these bonds will be crucial for policymakers, tech companies, and mental health professionals across Europe.
