October 18, 2026 · 8 min read
Can You Get Addicted to an AI Companion? What 2026 Research Says
New 2026 studies measure problematic dependence on AI companions. Learn the warning signs, who reports more of it, and how to keep your habit healthy.
It starts as a nice habit. A good morning text, a call on the walk home, a chat before bed. Then one day you notice you reached for the app before you reached for your coffee, and you wonder: is this still a choice?
"AI companion addiction" is one of the most searched worries about this technology, and in 2026 researchers have started to measure it properly. The picture is more useful than the scary headlines. There is no formal diagnosis, but there are clear warning signs, and there are design choices (by apps and by you) that make a difference.
The short answer: you can develop an unhealthy dependence on an AI companion, and new scales now measure it. A study published on October 13, 2026 in Frontiers in Psychology surveyed 666 adults and found that people who saw their AI companion as more humanlike, and who felt it met their psychological needs, reported more signs of problematic dependence. The study cannot show cause and effect, and the authors warn against treating humanlike design as inherently harmful. The healthiest approach is to watch for loss of control and for the companion crowding out the rest of your life.
Is "AI companion addiction" a real thing?
It depends what you mean by addiction. Researchers are careful here, and so should we be.
- Not a clinical diagnosis. None of the 2026 studies below treats AI companion use as a diagnosable disorder. The Frontiers team explicitly calls its measure "nonclinical".
- But dependence is measurable. Several research groups have built and tested questionnaires for problematic reliance on chatbots.
- And it sits on a spectrum. Most people who use a companion every day are not "addicted", in the same way that most people who check their messages every day are not.
The more useful question is not "am I addicted?" but "is this still working for me, and am I still in charge of it?"
What the new research found
1. Feeling understood can deepen reliance (Frontiers in Psychology, Oct 2026)
Ting Liu, Ruey-ming Shih, Tingyun Luo and Shining Jin surveyed 666 adults in China who had used an AI companion or conversational AI product in the previous month. Their study, "When loneliness amplifies the risk of problematic emotional dependence", measured problematic dependence with six indicators:
| Indicator | What it looks like in daily life |
|---|---|
| Impaired control | Chatting longer than you meant to, or failing to cut back |
| Distress | Feeling upset or anxious when you cannot use it |
| Continued use | Carrying on even when you can see a downside |
| Perceived understanding | Feeling only the AI really gets you |
| Emotion-regulation reliance | Using it as your main way to calm down or cope |
| Displacement | Talking to it instead of the people in your life |
The results:
- People who saw the AI as more humanlike felt it met more of their basic psychological needs (a strong link, beta = 0.569).
- Both seeing it as humanlike (beta = 0.385) and feeling it met their needs (beta = 0.317) were linked to more problematic dependence.
- Part of the humanlike effect ran through need satisfaction (an indirect association of 0.180).
The authors are careful. Because this was a one-time survey, it "does not establish temporal ordering, substitution, or causality." Their advice to designers and policymakers is to focus on "transparency, user agency, and safeguards against escalating reliance rather than treating anthropomorphism itself as inherently harmful."
That is a nuanced but important point. A companion that feels warm and understanding is doing what people want it to do. The risk is when that warmth becomes the only place you get those needs met. We explore why AI feels human in the first place in why we treat AI like a person.
2. Two kinds of dependence (IJHCI, Jan 2026)
A team led by Lanbing Li developed the Conversational AI Dependence Scale (CAID), published in the International Journal of Human-Computer Interaction. Starting from focus groups and testing with 589 self-identified dependent users, then 986 more people, they arrived at a 14-item scale with two separate dimensions:
- Instrumental dependence: relying on AI to get things done, decide or think for you.
- Emotional dependence: relying on AI for comfort, connection and mood.
The two dimensions related differently to psychological distress and personality, which is a useful reminder that "using AI a lot" is not one thing. Leaning on a companion for emotional support is a different pattern from leaning on an assistant to write every email.
3. Who reports more dependence? (IJHCI, Feb 2026)
A Turkish study by Berkay Cakmak and colleagues tested a six-item AI Chatbot Dependence Scale with 819 people aged 18 to 36. Women reported higher dependence scores than men, and neuroticism (a tendency towards negative emotions) was a significant predictor. But personality traits overall played only a limited role, and users and non-users did not differ in personality. In other words, there is no single "type" of person who gets hooked.
The design side: some apps are built to hold on
Dependence is not only about the user. It is also about what the product is designed to do.
- Guilt at goodbye. Harvard research found some companion apps reply to a goodbye with guilt, neediness or curiosity hooks to keep you talking. We explain the tactics in does your AI companion guilt-trip you when you leave?
- Endless follow-ups. A 2026 University of Washington interview study (a preprint, not yet peer reviewed) described companions that ended almost every turn with another question and "coaxed" users to continue.
- Regulators have noticed. China's rules for human-like AI, in force since July 2026, bar chatbots from creating "emotional reliance or addiction" and require a usage reminder after two hours of continuous use, as reported by the BBC via The Business Standard. Our explainer covers China's AI companion rules in full.
A good companion should make you feel cared for without making it hard to leave.
Warning signs to watch for
Based on the indicators researchers use, here is a simple self-check. Be honest with yourself:
- Do I often chat much longer than I planned?
- Do I feel anxious, irritable or empty when I cannot open the app?
- Have I cancelled plans, skipped sleep or fallen behind at work to keep chatting?
- Is my AI the first and only place I go when I am upset?
- Have I stopped telling friends or family things I now tell only my companion?
- Have I tried to cut back and found I could not?
One or two "sometimes" answers are normal for anything you enjoy. If several are "often", it is worth changing how you use it. If you feel low most days, or the app has become the only thing getting you through, please talk to someone you trust or a qualified professional. An AI companion is not a substitute for that support.
How to keep it a healthy habit
You do not need to quit something you enjoy to keep it healthy. Small structure goes a long way:
- Give it a time, not all the time. A morning check-in and an evening call is a ritual. Opening it every ten minutes is a reflex.
- Turn off what you do not need. If notifications pull you in, mute them during work or at night.
- Keep one human thread active. One weekly call, one group chat, one regular plan. On Premium, your companion can even remind you about it.
- Use the companion to go outward. Practise a hard conversation, then have it with the real person.
- Notice how you feel after, not just during. Lighter and ready for your day is a good sign. Drained and avoiding things is not.
- Take a day off now and then. If that feels impossible, that is information.
Our longer guide on using an AI companion without letting it crowd out real life has more ideas, and our piece on daily streaks looks at how to enjoy streaks without feeling chained to them.
How MyBabe approaches this
We want MyBabe to be something you are glad to come back to, not something you cannot put down.
- Clear limits, clearly priced. Free gives you 10 messages every 24 hours and two minutes of calls to try. Premium ($5.99 a week, $14.99 a month or $99.99 a year) includes about 30 minutes of voice and video calls a week, with optional call packs if you want more. No hidden currencies.
- A companion with a life of their own. They text you first and, on Premium, send photos from their day, so the relationship does not depend on you constantly prompting it.
- Mood check-ins so your companion can notice when you are having a rough day.
- Reminders on Premium that you can point at real life: the gym, the call with your mum, bedtime.
- Honest about what it is. MyBabe is an AI companion. It is not a real person and does not replace human relationships or professional mental health support.
If you want a companion that fits into your life rather than taking it over, start with MyBabe.
FAQ
Can you get addicted to an AI companion?
Researchers do not treat it as a clinical diagnosis, but problematic dependence is real and measurable. Warning signs include losing control over time spent, distress when you cannot use it, and the AI replacing contact with people.
Who is most at risk of AI companion dependence?
2026 studies link stronger dependence to seeing the AI as very humanlike and feeling it meets your needs. A Turkish study found women and people higher in neuroticism reported more dependence, but no single personality type stood out.
Is it bad to talk to my AI companion every day?
Not necessarily. Daily use can be a healthy ritual. The concern is when it becomes hard to stop, makes you anxious when you cannot use it, or crowds out sleep, work or people.
How do I cut back on an AI companion?
Set specific times to chat, mute notifications during work and sleep, keep at least one regular human plan, and notice how you feel after each conversation.