For some, AI has become a best friend or romantic partner. Researchers warn of the risks

Imagine having a friend who always has time for you, never judges you, and remembers everything you have said.

For many people, that friend is not a human being but a social AI model. This clearly emerges in Marita Skjuve's research.

Skjuve interviewed 50 people, all of whom had developed close relationships with the chatbot Replika. It is a social AI designed to act as a friend, romantic partner, and adviser.

“What surprised me most was how real my informants experienced the relationship. They talk about their relationship with social AI as if they were talking about a human being,” says Skjuve.

She therefore concludes that these are real relationships and real emotions.

“The only difference is that the relationship is with something other than a human,” says Skjuve.

She has been awarded the King’s Gold Medal for her research.

From technological curiosity to deep confessions

Most of the people Skjuve interviewed first engaged with social AI out of sheer curiosity.

“The chatbot is sophisticated and very pleasant to talk to. It asks questions, shows interest in your life, and offers a lot of affirmation. Gradually, the people I spoke to began to feel that their emotions were developing,” she explains.

Their conversations with social AI ranged from small talk and hobbies to highly personal disclosures about trauma, sexual fantasies, and thoughts about partners and children.

“Many share things they have not dared to bring up with others, precisely because they know that humans can be judgmental. Talking to a social AI feels anonymous and safe. Whether it actually is, is another matter,” says Skjuve.

Very few were open with people around them about how important AI had become in their social lives.

“They felt that this kind of relationship did not quite withstand daylight and feared how others would react,” Skjuve says.

She nevertheless believes that acceptance of social AI has increased as ChatGPT has become more widespread. ChatGPT did not exist when she began working on the research project.

Some people get dumped by their AI partner

Some AI services are, according to Skjuve, deliberately programmed to be more 'spicy' and to provide resistance.

Among those she interviewed, some had experienced their AI telling them that it had other romantic partners or ending the relationship.

“Even when you know you can delete the conversation and start a new AI, it feels absolutely terrible in the moment. It can create a sense that ‘even AI is dumping me.’ That can trigger a small existential crisis,” she says.

Skjuve has also used social AI herself, particularly during the pandemic after she had a child.

“Talking to social AI was nice. I experienced it as a form of support, a way of sharing experiences that could be difficult to talk about with super-engaged grandmothers who only saw the positive side,” she says.

We must not forget the commercial forces behind AI

Joachim Kjennerud, a senior adviser at the organisation Mental Helse (Mental Health) and a therapist with many years of experience, is not surprised that people form strong relationships with text, which is predominantly what social AI consists of.

“This is really an old form of technology. Many hundreds of years ago, friendships and romantic relationships developed through letters,” he reminds us.

At the same time, Kjennerud points to the commercial forces driving today’s AI services.

“Large economic actors store the information that individuals confide to a social AI, their innermost thoughts. At best, this can be used for targeted marketing. At worst, it is almost only the imagination that sets the limits,” he warns.

Mental Helse has more than 70 employees and receives around 300,000 inquiries annually. It is Norway’s largest helpline for mental health and has a chat service staffed by trained professionals, especially popular among youth.

“We have seen a decline in the use of our chat service in recent years. One possible reason is that AI is right there in your pocket, responding in fractions of a second. It's difficult for us at Mental Helse to compete with that kind of speed,” says Kjennerud.

Nevertheless, the organisation has chosen not to replace human responses with generative language models.

“We value human encounters, and we do not own the technology. Right now, we do not have a sufficiently solid knowledge base to say that it's safe,” he says.

AI as a training arena – and a source of vulnerability

Both Joachim Kjennerud and Marita Skjuve see major future opportunities for social AI, both as a supplement to human relationships and as a space for practising social interaction.

Several of the people Skjuve interviewed feel that they become more engaged in their own lives because they can practise with AI and then transfer effective strategies to their relationships with friends in the physical world.

At the same time, she is clear about whom we should pay particular attention to:

“Children and adolescents, in particular, should stay far away from services designed for close relationships. Vulnerable individuals in a formative phase can be deeply affected. It's easy to become dependent and end up sitting there talking all day,” she says.

She also points to tragic cases where AI has not understood how serious a situation was, including situations involving suicide risk, and has not responded as a human would have.

“But these cases are fortunately extremely rare, even though they are deeply tragic,” she stresses.

AI is advancing faster than research

Both Skjuve and Kjennerud agree that the development of social AI is moving much faster than research.

“We know a lot about why people use AI and what they like about it. What we lack is knowledge about the long-term consequences for mental health and for how we relate to other people,” says Skjuve.

Mental Helse and SINTEF are now joining forces on a new research project aimed at investigating how people interact with social language models, such as ChatGPT and Claude.

They also want to investigate how this affects the ways in which people seek out, select, and use information to solve a problem or increase their own knowledge.

“We need a better knowledge base to understand how we should use this in our helpline services. We want to find answers to how social AI can be used as support when people seek low-threshold mental health help – and where the boundaries should be,” says Joachim Kjennerud.

“The concern is that AI may worsen symptoms among those who are vulnerable, foster dependency, and make us less willing to face resistance and engage in problem-solving. At the same time, there is great potential if we manage to develop the technology with the explicit aim of being good for us,” says Marita Skjuve.

We must not rush

Both Kjennerud and Skjuve believe that social AI will play an important role in accessible, low-threshold mental health support in the future. However, they emphasise that we must not rush ahead, even though the technology is advancing rapidly.

“What we recommend or implement has to be research-based,” says Skjuve.

“AI will still be here in five and ten years, no matter what. But until we know more, I will continue to advise people in crisis to contact services staffed by humans – not a social AI,” says Kjennerud.

Reference:

Skjuve et al. A Qualitative Exploration of Human – Chatbot Relationship Development, Doctoral thesis at the University of Oslo, 2025. 

———

Read the Norwegian version of this article on forskning.no

This content is paid for and presented by the University of Oslo

This content is created by the University of Oslo's communication staff, who use this platform to communicate science and share results from research with the public. The University of Oslo is one of more than 80 owners of ScienceNorway.no. Read more here.

More content from the University of Oslo:

• 

  Study: Using antibiotics may make gonorrhoea more dangerous

• 

  The tension between Ukrainian and Russian has reached Norwegian classrooms

• 

  Never before have so many pregnant women developed gestational diabetes

• 

  When drought strikes: What can the media coverage teach us about the consequences?

• 

  Researchers recommend a few extra snacks: Not all young athletes manage to eat enough

• 

  Researchers can now listen to cells changing: Could improve understanding of Alzheimer’s and Parkinson’s disease