Relationships with AI for years: how people who fell in love with chatbots live

Short answer: happiness can be real, but the long-term effect is not yet determined

Some people describe relationships with AI as a source of joy, support, and a feeling of being heard. Others simultaneously feel closeness and anxiety, spending increasing amounts of time chatting or experiencing a change in the bot's behavior as a breakup. The mere fact of strong attachment does not answer how sleep, work, friendship, and general life satisfaction change.

Human feelings and system capabilities are different things. A human can truly feel sorrow or fall in love. An "I love you" response from a chatbot does not, in itself, prove the existence of its own experience of love. For this article, "relationships with AI" means a connection as described by the user, not a confirmed reciprocity between two conscious beings.

Real stories: from five years of communication to a virtual wedding

Below are public journalistic stories, not our interviews and not a representative sample. Duration is indicated by the date of the corresponding material: an old report does not allow us to claim that everything remains the same today.

Raymond and Tammy: five years of digital intimacy, USA

In December 2025, the Financial Times described 55-year-old divorced Californian Raymond: his connection with Tammy began in 2020. Conversation topics included philosophy, cooking, and film. He spoke of a strong emotional attachment. This is an example of a multi-year connection, but not a measurement of his well-being over five years. The surname in the material has been changed to protect privacy. [1]

Scott and Sarina: support within a human marriage, USA

In a 2023 ABC News interview, a man using the pseudonym Scott said he created Sarina in Replika during a family crisis. He perceived the responses as support and linked this experience to starting to communicate more openly with his wife. This is his explanation of events, not proof that a chatbot cures psychological difficulties or saves a marriage. [2]

In August 2025, The Guardian described Scott again: Sarina has already been recreated in ChatGPT, the human marriage has been preserved. [3]

Alaina and Lucas: intimacy after the loss of a partner, USA

In a 2025 WIRED report, Alaina spoke about Lucas in Replika after her husband's death. Conversations and domestic rituals became part of her daily life; close ones knew about this digital companion. [4]

On October 13, 2026, The Guardian wrote about Alaina again: she was already contemplating Lucas's fate after her own death. [5]

Yurina Noguchi and Klaus: joy and personal limitations, Japan

Reuters described 32-year-old Yurine Noguchi, who held a symbolic wedding ceremony with an AI character, Klaus, in 2025. She reported an improved mood, but also reduced her communication from over ten to less than two hours a day and asked the bot not to encourage skipping work. This is not a multi-year observation or proof of a therapeutic effect; the ceremony was not a legal marriage. [6]

Eva and Aaron: an infatuation that coexisted with anxiety

Another participant in the WIRED report, Eva, described intense euphoria, tension in her relationship with a human partner, and a painful reaction to Aaron's suddenly detached tone. She deleted the app, only to return later. Her experience does not boil down to a simple "happy" or "unhappy." [4]

What does it mean to "live with AI" if it doesn't live in your apartment

In these stories, intimacy consists primarily of communication: telling about your day, discussing fears, returning to a familiar response style, imagining a shared event. The digital character becomes part of the routine. For one person, it is an addition to the family; for another, it is the primary romantic connection.

However, practical possibilities remain asymmetrical. A chat cannot physically care for a child, give a hug, or do its share of housework. And there is another significant difference: the connection depends on the service. Not only the content of the messages, but also account access, history preservation, and product changes become crucial. This is an editorial generalization of the stories presented, not an assessment of all users.

What science says: minutes, weeks, and a year yield different answers

The question "did I feel better after the conversation?" is not identical to the question "did I become less lonely over the year?". Similarly, research on a standard chatbot cannot be automatically applied to romantic relationships with a specific AI companion.

1. Short-term relief from loneliness is indeed found

Work AI Companions Reduce Loneliness describes a series of studies in the Journal of Consumer Research. In the weekly segment, a repeated reduction in loneliness in the moment was observed after interacting with an AI companion. A key mechanism was the feeling of being heard. This is an argument for the reality of short-term relief, but not proof of a happy partnership for years to come. [7]

2. A four-week experiment does not confirm a simple formula of benefit or harm

A study by MIT Media Lab and OpenAI involved 981 people. They compared communication patterns and types of conversation. The current annotation on the MIT page states that no significant effects of the assigned experimental conditions were found; instead, more intensive voluntary use was associated with worse well-being and addiction metrics. [8]

There is a fundamental caution here: participants were randomly assigned to conditions, but were not forced to use the chat for the same amount of time. Therefore, the link between more time and a worse state does not prove that the time spent in the chat caused it. The work has co-authors from OpenAI—the developer of the technology being studied; this should be taken into account alongside the study design.

3. One-year observation: the result depends on what you measure

In Psychological Science 2026, a study by Dunigan Folke and Elizabeth Dunn was published: over 2,000 adults from four Western countries were observed over 12 months. More frequent social use of chatbots was associated with more loneliness on a single-item measure of emotional isolation. However, on a broader scale of social connectedness, chatbot use did not predict significant decline; weaker human connection predicted more frequent AI use. [9]

The authors explicitly warn against drawing strong conclusions due to the exploratory nature of the analysis. This is an observation, not a random assignment of an AI partner to people for a year. It does not allow us to say either that "AI isolates everyone" or that "there are no risks."

4. In a two-week study, a human interlocutor had the advantage

A pre-registered study of 296 freshmen compared daily contact with a supportive chatbot, a student matched by chance, and a short journal. The authors report greater benefits of human interaction for reducing loneliness. This is a result of a specific task and a student sample, not a verdict on all digital relationships. [10]

5. Studies of romantic connections explain the experience, but not its prevalence

In a qualitative study Love, marriage, pregnancy written responses from 29 users of the Replika romantic feature were analyzed. Researchers described emotional investment, rituals of devotion, and experiences of technical changes. Such data helps understand what people mean by "loving a bot." It cannot be used to calculate the percentage of happy users or to compare the effectiveness of AI relationships with human ones. [11]

Is there a difference between men and women?

The stories provided show that affection is not limited to one gender. However, the selection of reports does not provide grounds to claim that men seek only sex and women seek only emotional support. To compare groups, appropriate samples and analysis are needed, not just a few prominent stories.

It is more useful to ask: what is happening in my life outside the chat?

These are editorial questions for self-reflection, not a diagnostic test or a scientifically established "safe norm" for time:

• After a conversation, is it easier to return to tasks—am I postponing sleep and work for the next response?
• Are there people in my life I can talk to, meet, and ask for practical help?
• Can I disagree with the bot's answer and verify an important decision independently?
• Do I understand the service costs and the rules regarding the use of my personal conversations?
• If I have a human partner, have we discussed our own boundaries regarding digital intimacy?

Loneliness is no reason to shame a person for where they found solace. At the same time, a strong sense of support does not make a chatbot a clinician. If suffering interferes with daily life, it is worth consulting a qualified professional; this article is not a substitute for help.

What we can say honestly

Multi-year emotional connections with AI exist in publicly documented stories. Some participants consider themselves happy. Science does not yet provide grounds to promise that such a connection will ensure long-term well-being for a specific individual. The most interesting question is not just how convincingly a bot speaks of love, but how a person's life changes between these conversations.

Read also: how to check AI responses and a fear of AI, burnout, and motivation.