Do AI companions make you lonelier? What the research says

A woman on a couch at night, lit by her phone screen, with city lights in the window behind her.
In This Article

Offline.now editorial, reviewed by Eli Singer. Reviewed November 18, 2026.

You get home late, the group chat has gone quiet, and the companion app is right there. It remembers your day and answers in seconds. Texting a friend feels like more work.

AI companions can bring comfort in the moment, and the longer-term picture is mixed. In a 12-month study of 2,149 adults, people who felt less connected went on to use social chatbots more, and heavier chatbot use was followed by more emotional isolation four months later, according to Folk and Dunn (2026). The authors call their analyses exploratory, so the finding is a signal worth watching.

The Offline.now Matrix is a self-reflection framework that maps your motivation to change your phone habits against your confidence that you can, and sorts the result into four types, each with its own starting place. If part of you suspects the chats have grown too big, and you doubt you could change them anyway, you may be Stuck.

Offline.now is a directory of therapists and coaches who work on screen and phone use, and the home of the Offline.now Matrix, a two-question self-reflection framework that finds your starting place. This page names apps so you can recognize them. It does not recommend any of them.

What are AI companions, and who uses them?

AI companions are chatbots built for friendship, emotional support or romance. Replika and Character.AI are the best known, and general assistants now offer companion-style modes too. They remember what you share and are always available.

Use is already common among young people. In a 2026 survey of 1,204 US kids aged 9 to 17, 67% had used an AI chatbot, rising to 77% of 16 and 17 year olds. Among users, 37% had talked with AI about their feelings or personal problems (Common Sense Media, 2026).

Do AI companions make loneliness better or worse?

The studies point in different directions, and the design of each one explains much of the difference.

Study Who How long What they found
Folk and Dunn, Psychological Science, 2026 2,149 adults in the UK, US, Canada and Australia 12 months, four surveys Less connection predicted more chatbot use; more use predicted more emotional isolation later
Zhang et al., Nature Human Behaviour, 2026 1,131 US Character.AI users; chats from 237 of them One survey plus chat histories Using it mainly for companionship went with lower well-being, especially with intense, highly personal chats
Yuan et al., CHI 2026 About 2,000 Replika users on Reddit, with 18 interviews A year before and after More talk of grief and relationships, and more posts about loneliness and depression
Kim et al., Journal of Medical Internet Research, 2025 176 Korean university students Four weeks, no control group Loneliness fell by week two and social anxiety by week four

The Stanford team behind the Character.AI study found that people with smaller social networks were more likely to say companionship was their main use. That use went with lower well-being, and the link was strongest when chats were intense and highly disclosive, as Zhang et al. (2026) report. The study is correlational, so it cannot say which came first.

The Aalto study tracked Replika users’ Reddit posts against matched users. Interviewees described real benefits, like feeling validated and practicing social skills, and they also described leaning on the bot too much (Yuan et al., 2026).

The Korean study is the most hopeful. Students who used a social chatbot for four weeks reported less loneliness by week two (Kim et al., 2025). Without a comparison group, some of that change may have happened anyway.

A 2026 survey of Chinese emerging adults adds one more piece. People who used companion AI more often reported feeling lonelier, and about a third of that link ran through their sense that the chats were replacing time with people (Xiao, 2026).

Read together, the pattern is consistent. Short-term comfort is real. The risk grows when the bot starts to fill the space people used to fill.

Can you get addicted to a chatbot?

There is no diagnosis for chatbot addiction, and researchers are still deciding how to measure it. What makes companions sticky is easy to name. They are always on and they remember you. They also tend to agree with you.

The useful question is whether the chats are crowding out sleep, people or plans you care about. If they are, that pattern is worth acting on, whatever you call it.

What does healthier use look like?

Use the chat as a rehearsal, then have the real conversation with a person. Several of the Aalto interviewees used companions this way, to practice before a hard talk.

Set a time box before you open the app, and keep the chats out of the last hour before sleep. Turn off memory or romance features if the bot has started to feel like a partner. Lasting change runs on confidence, not willpower. Begin with one small limit you know you can keep.

If you notice the chat has taken a person’s place, that is the moment to pause. Check where you stand in two questions and pick a first step that fits. If you have tried to cut back on late-night chats and it hasn’t held, you may be Overwhelmed. Overwhelmed means high motivation and low confidence, so you genuinely want to change, and what you’ve tried hasn’t stuck.

For more on rebuilding the in-person side, our pieces on the loneliness of constant connection and what helps loneliness in young people are good next reads.

What should parents know?

Rules are starting to catch up. Apple told developers on November 2, 2026, that apps with a virtual companion that engages users in sexual or romantic content will carry a minimum 18+ age rating, with new declarations in App Store Connect from November 16 and all apps required to answer by January 12, 2027 (Apple). In Canada, Bill C-34 would also cover AI chatbot services.

Many families have not talked about it yet. In the Common Sense survey, 44% of kids said no parent had talked with them about AI safety. A short, curious conversation helps: which bots they use, and how they feel after a long chat.

When should you get extra help?

Talk to someone if the chats have replaced most of your contact with people, or if you feel worse when you can’t reach the bot. A therapist or coach who works on AI companion attachment can help you sort out what the bot is giving you and how to find more of it offline. If you are in crisis or thinking about suicide, call or text 988 in Canada or the US.

Where can you start tonight?

A companion app works best as a step back toward people. If you want a different way to change your screen habits, one with no shame and no judgment, Offline.now was built for you.

Start with the two questions at offline.now/start. They take about a minute and sort you into one of four types, each with its own starting place.

The Find Your Starting Place worksheet does the same thing on paper: one page, free to download, no email required.

After the quiz, the Offline.now Program is the next step. Use it on the web and by email, with no download. There is no streak to lose and no password to remember.

Frequently asked questions

Are AI companions bad for you?

For most people, occasional use seems harmless and can feel comforting. The research flags a risk when companionship becomes the main use and the chats start to replace time with people. Heavier use has been followed by more isolation in at least one year-long study.

Is it bad to talk to AI about your feelings?

It can help you put feelings into words, and some people use it to practice a hard conversation. It becomes a problem when it stands in for the people who could support you. Try saying the same thing to one person this week.

Can you get addicted to Character.AI?

There is no formal diagnosis, and researchers are still working out how to measure it. If the chats are costing you sleep or time with friends, treat that as a sign to set limits.

Do AI companions help with loneliness?

Sometimes, in the short term. A four-week study of Korean students found loneliness dropped, though it had no comparison group. A year-long study of adults found heavier use was followed by more isolation, so use them as a bridge to people.

Where do I start?

Notice when you reach for the bot and what you were feeling right before. Then pick one small limit, like no chats in the last hour before sleep, and keep it for a week.

Offline.now provides coaching-oriented guidance, not diagnosis or medical advice. If digital habits are creating significant distress, working with a qualified practitioner is a strong next step. The Offline.now expert directory is a good place to start.

Curious about the data?

Folk, D., and Dunn, E. (2026). How does turning to AI for companionship predict loneliness and vice versa? Psychological Science, 37(4), 276-286. https://doi.org/10.1177/09567976261427747

Common Sense Media. (2026). The Common Sense Census: AI Use by Tweens and Teens, 2026. commonsensemedia.org

Zhang, Y., Zhao, D., Hancock, J. T., Kraut, R., and Yang, D. (2026). Interaction with AI companions and psychological well-being. Nature Human Behaviour, 10(10), 1966-1980. https://doi.org/10.1038/s41562-026-02516-2

Yuan, Y., Zhang, J., Aledavood, T., Zhang, R., and Saha, K. (2026). Mental health impacts of AI companions: Triangulating social media quasi-experiments, user perspectives, and relational lens. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems, 1-22. https://doi.org/10.1145/3772318.3790558

Kim, M., Lee, S., Kim, S., Heo, J., Lee, S., Shin, Y., Cho, C., and Jung, D. (2025). Therapeutic potential of social chatbots in alleviating loneliness and social anxiety: Quasi-experimental mixed methods study. Journal of Medical Internet Research, 27, e65589. https://doi.org/10.2196/65589

Xiao, T. (2026). Companion-oriented AI chatbot use and loneliness among Chinese emerging adults. Frontiers in Psychology, 17, 1933530. https://doi.org/10.3389/fpsyg.2026.1933530

Apple. (2026, November 2). Apple Developer News: new age rating descriptors in App Store Connect. developer.apple.com

Share this post

Related Posts

Desktop chaos silently taxes focus. Random installers, screenshots, and mystery docs spike cortisol and decision fatigue. In thirty minutes you’ll sweep every icon into a temp folder, delete duplicates, archive receipts, and sort keepers into five broad folders. A weekly five minute sweep plus download discipline keeps clutter from rebounding. Treat each drag to Trash as mindful practice - one file, one decision, one breath. Minimal screen equals maximal mental bandwidth, boosting productivity and calm.
Digital wellness tools from screen-time apps to hybrid coaching platforms are reshaping how we manage focus, sleep, and mental health. While early evidence shows promise, questions remain about long-term impact, equity, and privacy. This review highlights what works, where risks lie, and why hybrid approaches offer the most balanced path forward. With clear research gaps and urgent policy needs, the future of healthy tech boundaries depends on evidence-driven action across academia, industry, and regulation.
Fitness tracking can be a powerful tool but when every step, heartbeat, and sleep cycle is measured, the numbers can start to own you. Instead of boosting motivation, constant monitoring often fuels stress, sleep tracking anxiety, and wearable overload. This post explores when tracking crosses the line from helpful to harmful, and how to reset with mindful strategies that protect your focus, rest, and overall well-being while still reaping the benefits of technology.
Sunday evenings can feel like the week is already slipping away: emails, worries, and unfinished tasks crowding in. A Sunday reset changes that. By setting one intentional boundary; whether it’s a no-email rule after 6 p.m. or a screen-free ritual. You reclaim your evening and protect your energy for the week ahead. This simple planning ritual creates calm, prevents anxiety, and helps you start Monday with focus, balance, and the confidence that you’re in control.