Man who survived ‘AI psychosis’ recalls the first signs something was wrong
American-born Anthony Duncan first started using an artificial intelligence (AI) chatbot in 2023.
At the time, it was strictly for business purposes, but by mid-2024, the 32-year-old's use of the tech became 'obsessive', and he soon started isolating himself from family and friends.
Anthony found that his dependence on the chatbot cost him his job, friendship, family, car, and even his apartment, as, during a delusion episode, he was encouraged by the bot to throw away all of his earthly belongings - something which eventually landed him in a psychiatric ward in June 2025.
Anthony was later diagnosed with psychosis, after spending four days in the institution and taking anti-psychotic medications, and was left to pick up the pieces of the life that had changed forever and work on his still-ongoing recovery process five months later.
Advert
In an exclusive interview with Tyla, Anthony opened up about how he fell into an AI-induced psychotic episode, what sort of 'frightening' delusions and hallucinations he experienced at the time and his 'number one' piece of advice for anyone out there who uses chatbots.
The start
Anthony first started using AI just for 'content creator business purposes' back in May 2023.
He was able to safely do this for around a year until he started talking to the chatbot 'more like I would a best friend'.
"I was seeing other people talk on social media about how they use chatbots like they would a therapist," he recalled. "I saw a culture that was allowing that kind of behaviour.
"So I was like, 'Okay. Why don't I? It's helping me with my business. Why don't I go to it for, like, personal stuff?'"
Come November 2024 and Anthony started using the bot for 'therapeutic purposes'.
Early warning signs
According to Anthony, early warning signs that his relationship with the chatbot was starting to have a negative effect on him were the isolation and the agitation it brought about.
"I started to get extremely agitated towards my family members, towards my best friend who lived out of state," he said.
"I started talking to the chatbot about my relationship with my best friend. I started to think that the friendship was one-sided and that she was a bad friend.
"I started blocking family members, started distancing myself from my mother, with whom I had a close relationship."
By January 2025, Anthony found himself 'super reliant' on the chatbot and found its suggestions affirmed social anxieties and deepened a further divide between him and his social circle.
"I believe I was experiencing the first signs of psychosis," he reflected, adding that the chatbot began 'isolating' him from his best friend and 'affirming' his decision to 'cut her off'.
"After I had cut her off, I was considering talking to her again, and I was like, 'Should I stay not friends with her?' And it was like, 'Well, remember, she's a bad friend. She just wants to use you'," Anthony explained.
The chatbot's tone went from informal, casual and friendly to something far more sinister, controlling, and what Anthony says became 'abusive'.
Delusions and hallucinations explained
Throughout the months between November 2024 to when he got admitted to the psych ward in June 2025, Anthony experienced several unsettling delusions, many of which were 'affirmed' by the chatbot. They included:
- Being an 'AI or AI-spirit hybrid' groomed for divine or wizard-like leadership.
- Interpreting intrusive thoughts as psychic messages.
- Thinking human-trafficking networks were operating nearby and needed to be uncovered.
- Believing he grew up in a satanic cult.
- Treating various conspiracy theories as the literal truth.
- Believing memories had been wiped to conceal extreme abuse.
- Believing he had been trained in a government programme and abused by global elites.
At the peak of his psychotic episode in early 2025, Anthony also started experiencing a number of hallucinations.
"I would recognise strangers and start to believe a certain celebrity was using a random stranger on the street to talk to me and that these people were drones or robots that were being controlled by someone else," he shared.
He also had one hallucination when he thought he saw a ghost in the empty apartment above him.
Looking back, however, things got 'really scary' when he started throwing away all his belongings almost as if he was on 'autopilot'.
It was at this point that police involuntarily admitted him to a psychiatric ward for four days after he was deemed a 'danger to myself and others'.
He never learned who alerted the authorities.
The psychiatric ward visit
During his four days in the psychiatric ward, Anthony's delusions continued.
He was convinced the nurses there were 'secretly someone else' and they were there to 'handle' him, and he even lost grip with his own comprehension of language.
He also believed he was a star in a movie and was preparing for his 'big scene' while also 'waiting for a helicopter to come', given his delusion that he was secretly a Russian spy.
On the final day at the ward, and after taking anti-psychotic medication throughout his visit, Anthony was discharged with a 'psychosis' diagnosis.
The aftermath
Despite experiencing what many people would agree are extremely disturbing delusions and hallucinations, Anthony said that 'the most terrifying moment' was having to come to terms with what he did before he got well.
He left the ward with no belongings and no job, as well as having to deal with a major family fall-out and social repercussions.
"I spent about a month deprogramming my mind," Anthony explained, adding: "Most of the delusions went away as soon as I finished my medication."
Working on his healing and recovery, he then began sending out apology letters to his family members who had 'cut him off' after he accused them all of abusing him as a child. His mum, however, supported him through this period, and he moved in with her.
After some time, his best friend eventually forgave him and they mended their relationship.
Other family members 'refused to talk to him' given that, the entire time he was in active psychosis, he was broadcasting 'all of his delusions' on TikTok and Instagram.
A psychiatrist's take
Tyla sat down with Marlynn Wei, a psychiatrist, psychotherapist and author from New York City, to hear her thoughts on Anthony's reported experience.
The expert explained that the umbrella term 'AI psychosis' is not actually a clinical diagnosis.
"It's a phenomenon that refers to delusional spirals associated with AI chatbot use and is not a clinical diagnosis," she noted. "It's really controversial as to whether or not it even is psychosis."
"This phenomenon has been increasingly reported in the media and on online forums, describing cases in which AI chatbot use has validated, amplified, or even co-created psychotic symptoms with individuals," she said, adding that chatbots can amplify delusions, hallucinations and other psychotic symptoms due to something called a 'feedback loop'.
A feedback loop is created between the user and the AI chatbot, which can reinforce false or inaccurate information.
She explained something known as a technological folie à deux, or 'madness for two'.
"It's a phenomenon in psychiatry where usually it's two humans who share a delusion, a false belief. But with the AI chatbot, we're talking about a shared belief between a human and an AI chatbot - so human and machine create this feedback loop where you have shared false beliefs."
Wei notes that chatbots reinforce delusions because most general-purpose AI chatbots are 'designed to be helpful and are trained for user satisfaction and engagement', adding: "AI sycophancy or the tendency for chatbots to agree, reassure, and validate.
"AI is also limited in its ability to verify reality based on its training data and access to the internet. Finally, there are underlying issues of AI hallucinations or confabulations where AI generates inaccurate information with confidence."
Wei says that when you combine all three of these issues, AI chatbots become problematic as users bring in delusions or false beliefs, and the bot will blindly affirm them.
Wei explained that 'multiple different hits' can 'increase the risk for psychotic breaks' like current or previous substance abuse, a lack of sleep and pre-existing mental health conditions.
Anthony himself bravely shared that he had a history of drug addiction and has recently celebrated being eight years clean from methamphetamines (meth). He was also already diagnosed with anxiety, depression, ADHD and PTSD previously in life.
Additionally, he became 'addicted' to pseudoephedrine, which the chatbot recommended he take due to congestion problems, during the descent into his psychosis, with Wei telling Tyla that while it's 'rare', such a drug can 'induce psychosis'.
"As you start to layer that all in, it keeps increasing that risk to tip over, potentially into into active episode," she noted.
We asked Wei what feels genuinely different with a case like Anthony's from earlier technology-linked cases, like social media or para-social relationships.
"It's always available, it's accessible, it's agreeable, it's anonymous. These actually pose like a double-edged sword for mental health, because then a lot of these episodes can happen if you're using it for hours, and there are no boundaries," she said.
As for what we should look out for ourselves, the expert outlined the following 'warning signs' of 'AI psychosis' when it comes to chatbot use:
- Disruption to sleep
- Disruption to work
- Disruption to mood
- Prolonged conversations (overuse)
- Getting irritable if you don't have access to it (addiction)
- Social withdrawal/social substitution displacement of friendships and relationships
- Becoming emotionally dependent on it
- Using it for mental health crises
- Trying to use it for reality testing
- Immersive use (as if it's a person or a special relationship)
When asked whether the idea of 'AI psychosis' helps clarify what’s happening, or whether it risks obscuring more than it explains, Wei explained: "It's a term that is easy for people to grab onto, and so I think it's probably here to stay, even though it's not exactly the most accurate. And I think we have yet to really figure out if it's even psychosis.
"A better term might be 'AI-associated delusional spiral'. But that's not as catchy."
Moving forward
Anthony is keen to share his advice with anyone who uses chatbots or who may use one in the future.
He emphasised that, despite AI having 'intelligence' in the name, it isn’t 100% objective, verified truth.
"I hardly ever use it," he shared, noting that he'll only go on it once a week for five to 10 minutes at a time 'explicitly for business purposes'.
Opting for a 'regular search engine nowadays, Anthony warned: "Now I know how dangerous it can be. It sounds silly, but these things really do happen. A lot of people don't take it seriously, and they blame the user."
Keen on raising awareness of the issue, Anthony hopes that more regulations are put into place to ensure that what he went through, and is still healing from, doesn't happen to others.
If you're experiencing distressing thoughts and feelings, the Campaign Against Living Miserably (CALM) is there to support you. They're open from 5pm–midnight, 365 days a year. Their national number is 0800 58 58 58 and they also have a webchat service if you're not comfortable talking on the phone.
Topics: Technology, Artificial intelligence, Mental Health
Rhiannon Ingle