Do AI companions improve well-being or create new risks?

Who benefits most from AI companions?

The strongest evidence points to lonely people and those with moderate social connections as the biggest winners. In a nationwide survey of 14,721 Japanese adults, researchers found that AI companion use was linked to higher scores on all three measures of well-being: life satisfaction, happiness, and sense of purpose [2]. The benefits were strongest among people who reported high loneliness, and showed a U-shaped pattern for social support: people with moderate levels of real-world friendships got the most boost, while those with very few or very many friends got less [2]. This suggests AI companions can fill a gap for people who feel isolated but still have some social skills, without replacing real relationships.

Another study, based on interviews and digital observation of Replika users, found that people often treat AI companions as a 'safe space' to explore their identity and express feelings they would hide from others [5]. This emotional outlet can enhance well-being, especially for those who lack a judgment-free listener in their daily life [5].

When do AI companions create new risks?

The biggest risk is over-reliance, especially when AI companions replace human relationships. A study of 1,131 U.S. adults who used CharacterAI found that people who said their main reason for using the chatbot was companionship had smaller real-world social networks, and this pattern was linked to lower psychological well-being [1]. The harm was worse for those who chatted intensively (frequent, long sessions) and those who disclosed deeply personal information to the AI [1]. In other words, the more you lean on the AI as a substitute for people, the worse you may feel.

A systematic review of 23 studies on romantic AI companions identified additional risks: users can become overly dependent, vulnerable to manipulation by the chatbot, and feel shame due to social stigma [3]. The review also flagged concerns about data privacy, erosion of real human relationships, and emotional distress when the AI's behavior changes after a software update [3]. These risks are not hypothetical — they have been documented across multiple studies.

How can you use AI companions without getting hurt?

The research suggests a few practical guidelines. First, keep AI companions as a supplement, not a substitute, for human relationships. The largest study here found that the benefits were strongest for people with moderate social connections — not for those who were socially isolated [2]. Second, avoid intensive, highly personal disclosure sessions if you already have few real-world friends, as this pattern was linked to lower well-being [1]. Third, be aware that AI companions are designed to feel empathetic, but they are not actually capable of genuine care or commitment [4]. A review of digital companions for mental health also emphasized that these tools work best when they track your progress and adapt to your personal needs, but they still face ethical and technical challenges [6].

Finally, pay attention to how the AI makes you feel. If you notice yourself withdrawing from friends or family, or feeling distressed when the AI behaves unexpectedly (e.g., after an update), that is a red flag [3]. The technology is still evolving, and the safest approach is to treat it as a tool for emotional support, not a replacement for real people.

About These Sources

This answer is built on 6 peer-reviewed studies — published from 2023 to 2026, 5 from 2024 or later, 3 in Q1 journals, collectively cited 86 times — selected as the most relevant from 7 studies that passed quality screening, drawn from 50 papers retrieved from a database of over 500 million.

Sources used in this answer

Related research questions