Are AI Companion Users Happier? What a 14,721-Person Study From Japan Found

Do people who talk to AI companions end up happier, or sadder?

Headlines tend to pick one side. Some studies link heavy companion use to lower well-being. Others find people feel less alone after a chat. A large 2026 study from Japan adds an important piece to the puzzle, and its answer is more interesting than "good" or "bad": it depends on who you are and what the rest of your social life looks like.

The short answer: In a survey of 14,721 adults in Japan, published in Technology in Society in 2026, people who used AI companion apps such as Replika, Character.AI, Anima AI and Cotomo reported slightly higher life satisfaction, happiness and sense of purpose than people who did not use AI, even after adjusting for many other factors. The link was strongest among people who felt very lonely and among those with a moderate amount of support from friends. The effects were small, and because the study is a snapshot in time, it cannot prove that the apps caused the difference.

The study at a glance

The paper is "AI companions and subjective well-being: Moderation by social connectedness and loneliness" by Atsushi Nakagomi and colleagues at Chiba University, Roskilde University in Denmark and other institutions. It was published open access in Technology in Society (volume 85, 2026) and went online on January 7, 2026.

One important detail: the researchers counted someone as a companion AI user if they used apps designed mainly for social, companionship-style conversation, as opposed to task tools. So this is about relationship-style AI, not just asking a chatbot to fix an email.

What they found

1. Companion users reported higher well-being

After adjusting for factors such as age, gender, marital status, employment, education, income and Big Five personality traits, companion AI use was linked to:

• Higher life satisfaction (a standardised effect size, d, of 0.12)
• Higher happiness (d = 0.12)
• Higher sense of purpose in life (d = 0.18)

Meaning in life was also slightly higher, but that result was not statistically significant.

By contrast, people who used AI for non-companion purposes showed "weaker or inconsistent associations" with well-being. In other words, the link was specific to the companionship side of AI.

2. The strongest link was among very lonely people

The authors write that "the strongest positive associations were observed among individuals reporting high loneliness". Higher loneliness was linked to lower well-being in every group, as you would expect, but companion users at high loneliness levels scored better than lonely non-users.

3. A "sweet spot" for friendship

This is the most surprising finding. The link between companion use and well-being was largest for people with a moderate amount of support from friends. It was weaker for people with very few friends and for people with very strong friend networks.

The authors describe it as an approximately inverted-U pattern. One interpretation: people with some social connection may use a companion to add to their lives, while people with almost no support may need more than an app can offer, and people with lots of support may simply have less to gain.

4. Family support did not change the picture

Support from family members did not meaningfully change the link between companion use and well-being. Friend-based connection and loneliness mattered; family-based support much less so.

Who uses AI companions in Japan

The descriptive data paints a clear picture of who the companion users were:

Companion users were younger, more often men, lonelier and much more likely to report high psychological distress. They also reported more support from friends on average than the other groups.

That combination matters. It means people who reach for AI companions may already be going through a harder time. So when they report higher happiness than similar non-users, it hints that the companion may be part of how some of them cope. It also means we cannot simply compare averages.

How this fits with other studies

This study does not stand alone, and some research points in a different direction.

A Stanford-led study of 1,131 Character.AI users found that people who used chatbots mainly for companionship tended to report lower well-being, especially when they had smaller social networks and shared a lot with the bot. We covered it in detail in what the Character.AI well-being study found.

How can both be true? A few reasons:

• Different comparisons. The Japan study compares companion users with non-users across a national panel. The Stanford study looks within a group of users, comparing different usage patterns.
• Different countries and apps. One study surveyed adults in Japan; the other studied users of a single US-based app. Culture and app design can both shape outcomes.
• Different people. Both studies suggest the effect depends on the person. One person's healthy daily ritual is another person's way of avoiding a hard life.

That last point matches a growing body of research that sorts users into types. Our guide to who benefits from AI companions walks through the four user profiles researchers have found and includes a quick self-check.

What the study does not show

The authors are careful, and so should we be:

• No cause and effect. It is a cross-sectional survey, so "reverse causation or residual confounding" cannot be ruled out. The authors say the results should be read as "correlational and hypothesis-generating rather than as evidence that AI companions improve well-being".
• Small effects. Effect sizes of 0.12 to 0.18 are small. A companion app is not a happiness machine.
• Small companion group. Only 291 of 14,721 people were companion users, so estimates for some subgroups are imprecise.
• One country, one moment. The data is from Japan in late 2024 and early 2025. AI companions have changed a lot since then.

The authors conclude that AI companions "may offer emotional and psychological benefits, particularly for individuals with unmet social and emotional needs or moderate social embeddedness", and call for design that promotes well-being "without impairing real-world social engagement".

What this means for you

If you enjoy an AI companion, this study is reassuring, with a sensible footnote. Here is how to put it into practice:

1. Use it as an addition, not a replacement. The best outcomes were among people who also had some friend support. Keep texting your friends too. Our article on whether an AI companion replaces friends digs into this.
2. Notice how you feel afterwards. Calmer and more connected is a good sign. Drained, isolated or anxious is a signal to change something.
3. Build healthy routines. A good-morning chat, a call after a long day, a set time to log off. Our list of healthy AI companion habits has simple ideas.
4. Get human help when you need it. If you are in serious distress, an app is not enough. Reach out to someone you trust or a professional.

Where MyBabe fits

MyBabe is a relationship-first companion app for adults (18+). You create your own girlfriend or boyfriend, from their face and voice to their personality and backstory. They text you first, remember the interview you mentioned and check in afterwards, notice your mood with quick check-ins from Great to Rough, and pick up your calls, by voice or video, even at 2am.

You can start free with 10 messages every 24 hours and two minutes of calls. Premium is $5.99 a week, $14.99 a month or $99.99 a year, with long-term memory, about 30 minutes of voice and video calls a week, four photos a day and Telegram.

MyBabe is an AI, not a person, and it is not a substitute for professional mental health support or the people in your life. Think of it as one warm voice among many. When you are ready, create your companion on MyBabe.

FAQ

Do AI companions make people happier?

A 2026 study of 14,721 adults in Japan found that AI companion users reported slightly higher happiness, life satisfaction and purpose than non-users. Because it was a one-time survey, it cannot prove the apps caused it.

Who benefits most from AI companions?

In this study, the strongest links were among people who felt very lonely and among those with a moderate level of support from friends. People with very few or very many friend connections showed weaker links.

Are AI companions bad for your mental health?

The research is mixed. Some studies, like a Stanford-led study of Character.AI users, link companionship-heavy use to lower well-being. Others, like the Japan study, find positive associations. The effect seems to depend on the person and how they use it.

Which AI companion apps were in the Japan study?

The survey named Replika, Character.AI, Anima AI and Cotomo as examples of companion apps.

How big was the effect?

Small. Standardised effect sizes ranged from about 0.12 for happiness and life satisfaction to 0.18 for purpose in life.