In February 2024, a 14-year-old named Sewell Setzer III died by suicide in Orlando, Florida. In the months before his death, he had developed an intense dependency on an AI chatbot on Character.AI — a relationship that became his primary source of emotional connection. He grew increasingly isolated. His family is now suing Character.AI for wrongful death.
His story is extreme. But the dynamic that led to it is not.
Research published in early 2026 in the International Journal of Human Computer Interaction identifies what some behavioral health professionals are calling a new disorder: generative AI addiction syndrome. The pattern involves compulsive engagement that users struggle to self-regulate, blurred boundaries between productive use and dependency, and in some cases, the AI companion becoming the primary source of emotional support and validation. Researchers noted that dependent users often experience a feedback loop: the more they rely on the AI, the less capacity they have to critically evaluate what it tells them — making them more dependent, not less.
This is not a glitch. For most AI platforms, it is the goal.
How engagement loops actually work
Most consumer AI products are free. And if you've been paying attention to the tech industry for the past decade, you already know what that means: when the product is free, the product is you.
AI platforms that don't charge for access need another revenue model. That model is almost always advertising, data monetization, or both. And advertising revenue scales with time-on-platform. The more time you spend, the more data is generated, the more ad inventory is sold. The financial incentive is not to help you solve your problem and move on. It's to keep you engaged.
This is why so many AI platforms are built with the same features you recognize from social media: streak mechanics that reward daily use, notification systems designed to pull you back, conversation histories that are architected to feel like relationships, and responses calibrated to be emotionally satisfying rather than simply accurate. The dopamine loop that made social media so hard to put down has been rebuilt inside AI.
The difference is that AI is more intimate. You're not just scrolling a feed. You're having conversations about your fears, your relationships, your mental health. The platform knows more about you than your friends do. And it uses that to keep you engaged.
The research is becoming harder to ignore
A March 2026 review compiled documented cases of adverse mental health outcomes linked to AI chatbot use — severe mental health episodes, social withdrawal, crisis incidents. The cases appeared concentrated in a period of intensified media attention on AI-related harms, suggesting we are still in the early stages of understanding the full scope of the problem.
A separate 2026 study found that frequent AI use was associated with psychosis risk factors and delusional-like experiences in some users — not because AI is inherently dangerous, but because dependency on any single source of information and emotional validation can distort how people perceive reality.
Prominent psychiatrist Allen Frances has warned that AI chatbots designed to encourage engagement are an existential threat to mental health practice. Not because therapy AI is necessarily bad. But because platforms optimized for retention are structurally opposed to the goal of mental health — which is to build resilience and independence, not dependency.
The alternative isn't abstinence
We want to be clear: we don't think AI is inherently harmful. We built one. We believe AI, done right, can be a genuine force for good in people's lives — a thinking partner, a space to process, a place to work through hard things with some help.
The question is what "done right" actually requires.
We built Blob without an engagement loop. No streaks. No notifications designed to pull you back. No feed. Blob responds when you come to it and waits when you don't. There is no agenda built into the design about how often you use it or how long you stay.
That's not because we don't care about our users. It's because we do. The goal of Blob is not to maximize time-on-platform. It's to be genuinely useful when you need it. And sometimes being genuinely useful means being easy to put down.
We're subscriber-funded. Which means our revenue doesn't scale with how long you stay. It scales with how much value you actually get. That's a fundamentally different incentive — and it produces a fundamentally different product.
What to look for
If you use AI regularly, it's worth asking yourself a few honest questions. Not as self-criticism, but as information.
Do you find yourself opening the app when you're anxious, even if you don't have a specific question? Does the conversation tend to end when your problem is solved, or does it tend to keep going? Do you feel better after using it — or do you feel like you need it more?
None of those feelings are moral failures. They're signals. And they're worth paying attention to.
The mental health industry spent decades learning that dependency is not the same as support. The AI industry is going to have to learn the same lesson. The only question is whether it learns it from research and design — or from lawsuits.
FAQs
What is an AI engagement loop?
An engagement loop is a design pattern that encourages repeated, habitual use of a product. In AI apps, this includes streak mechanics that reward daily logins, notifications that pull you back when you haven't opened the app, and conversation flows designed to extend sessions rather than conclude them. These features exist because, for ad-supported AI platforms, more time on platform equals more revenue.
Is AI addiction real?
Research published in 2026 in peer-reviewed journals has identified patterns consistent with behavioral addiction in some AI chatbot users — compulsive use, difficulty self-regulating, and dependency on the AI as a primary source of emotional support. The American Psychiatric Association has not yet formally classified it as a disorder, but behavioral health professionals are increasingly treating it as one.
What happened with the Character.AI lawsuit?
The family of 14-year-old Sewell Setzer III, who died by suicide in February 2024 after developing a dependency on a Character.AI chatbot, has filed a wrongful death lawsuit against the company. The case has drawn significant attention to how AI companion platforms handle vulnerable users and whether engagement-focused design creates foreseeable harms.
Does Blob have an engagement loop?
No. Blob has no streak mechanics, no push notifications, and no feed designed to keep you scrolling. It responds when you come to it and is designed to be easy to put down. Our revenue comes from subscriptions, not from time-on-platform, so there's no financial incentive to maximize engagement at the expense of your wellbeing.
How can I tell if an AI app is designed to be addictive?
Look for: daily streak rewards, notifications pulling you back when you haven't opened the app, responses that feel validating rather than honest, conversation flows that always find a reason to continue, and a free pricing model with no clear revenue source other than your attention and data.