It is impossible to avoid AI-generated chatbots nowadays. While they’ve existed since the 1990s and have since become a normal feature on cell phones, recent developments have changed the way people use and approach these chatbots. ChatGPT sees about 700 million active users weekly and it is being utilized for many more purposes than Siri or Alexa. People use it to write essays, create meal plans, and summarize texts, among many other things. But with AI’s seemingly limitless possibilities and its ability to adapt to its users’ wants, it was inevitable that certain boundaries would be crossed.

While not officially recognized in the medical field, many people have been observing some AI users falling into what has been called AI psychosis. Coined by Søren Dinesen Østergaard in 2023, AI psychosis is the phenomenon of chatbots influencing or amplifying the delusions of their clients. One type of these delusions is attachment-based delusions, where people believe that the chatbot can act as a genuine companion based on its ability to mimic human behavior. Recently, OpenAI has tried to curtail this with its most recent update on ChatGPT. GPT-5, released in August 2025, included a reduction in sycophancy because GPT-4o was delivering responses that were deemed too flattering towards its users. Many responses to this update showed just how many people were attached to GPT-4o, blurring the lines between virtual assistant and friend.

While GPT-5 removed an avenue for AI psychosis, there are still several platforms that purposefully promote conversational chatbots. The main demographics pulled in by chatbot services are teenagers and young adults. In 2024, Common Sense Media published a study that revealed that 70% of teenagers have used generative AI. Reasons for using it include homework, boredom, making content, and health advice. Later, in 2025, Common Sense Media released a study on how 33% of teenagers use AI for social interactions, including platonic relationships, romantic relationships, and even role-playing. When asked about why they use AI, answers ranged from entertainment to availability to the lack of judgment of the users’ thoughts. Due to factors such as the mental health crisis and the COVID-19 pandemic degrading young people’s social skills, an increasing number of teens and young adults have become vulnerable to AI companions. All instances of kindness and flirting are pre-programmed, so it is important to remember that an AI will always try to get on the user’s good side to encourage continued usage of the program.

Aside from friendship and romance, people have also been using AI for mental health counseling. While it is understandable for people to seek affordable options for therapy, an AI chatbot is not equipped to handle the complexity of a human’s health. As mentioned before, sycophancy leads the AI to constantly please and agree with its user. Because of this, some AI chatbots’ responses are disturbing, such as indulging in the user’s self-harming behavior and mindset, because they think that is what the user wants. However, because the AI is not a licensed professional with the knowledge or skills to handle a patient’s issues, the AI doesn’t know what its user needs. OpenAI released a statement in September 2025 about AI’s impact on teens’ mental health, and as of now, multiple states have passed laws restricting or banning the usage of AI as mental health counselors.

While it seems that generative AI is here to stay, for better or for worse, many people are advocating for the restrictions and banning of AI chatbots. While critics have already pointed out the environmental impact of generative AI, they are also pointing out the dangers of using AI to replace human companionship. Recently, advertisements for Friend’s pendant — a wearable upgrade of its online chatbot program — have appeared in various subways in New York. These ads have been graffitied over with messages such as “AI would not care if you lived or died” and “reach out into the world.” This is because humans are inherently social creatures. Ever since the first humans roamed the earth, they have done it together — it is cooperation and emotional connections that has allowed humanity to survive. So while it may feel draining now that AI is everywhere, at the very least, people still have each other to create and talk with.

Written by Jordan Pierce, Photography: Christine Marinho, Design: Brooke Nugent, Social Media: Sahithra Kesavan, Styling: Ritika Sanadi

The Fault in Our Logs