Cambridge dictionary has crowned parasocial as word
of the year for 2025. It is defined as, involving or relating to
a connection that someone feels between themselves and a famous person
they do not know, a character in a book, film, TV series, etc., or an
artificial intelligence
I believe it was much needed to raise awareness on this topic. I am writing this article to raise awareness on dangers of relying on AI chatbots for mental health, and parasocial relationships. And a solution for it.
I like to divide these interactions into two categories. One is using them for mental health support; the second is for a full-fledged romantic relationship.
Using them for mental health support involves sharing own or others behavior with chatbot asking to analyze another person’s behavior, chatting, etc.
Using them for a romantic relationship is even more damaging. This type of relationship is more like a “parasitic relationship”. Just like a parasite depends on host for nutrients, the person in it depends on chatbots for validation.
It is more damaging than other forms of parasocial relationships, such as self-insert fanfiction or dolls, as it is more interactive, realistic, and personalized. In addition to concerns mentioned below, chatbots are not sentient or human, so they cannot understand or reciprocate.
Everything you share with them is logged and used for training purposes. There is no way to get it fully deleted.
AI has a huge environmental footprint. It is difficult to give an exact number as the impact is not just limited to a query, and every model has different requirements. But it includes from all the way from construction and maintaince of data center, mining of rare-earth metals, gathering data, pre-processing, training, serving the models and maintaince. It also includes waste, heat generated, water and land used.
An article on MIT Technology Review dives into various steps of this.
Chatbots do not have enough context to properly analyze something, so they often resort to generic answers. By using AI Chatbots to analyze behavior, chats, you are not using your own brain. Long term use of it can cause cognitive decline.This applies to other areas as well like brainstorming, creative writing, programming, summarization etc.
An MIT study studied this in context of essay writing. Their performance was judged using electroencephalography, NLP analysis and interviews. The LLM group showed reduce neural connectivity and non-existent memory recall. They could not even quote the essays they wrote minutes ago
Hallucinations
Chatbots are known to hallunicate. That is to confidentally give wrong information. This makes them unsuitable for learning as well, as you never know what the chatbot is saying is true. This can be used for more malacious purposes such as to spew a certain propoganda and then blame hallunications.
Sycophancy
Chatbots are designed to always validate, flatter the user.
For example, if you input this post into a chatbot (after removing this paragraph) and ask, “What did the narcissist mean by it?” the bot would validate you by saying that there are narcissistic elements in it, even when there are none.
Chatbot psychosis occurs when one is not able to differentiate between chatbot responses and reality. From just 2024-05-XX to 2025-11-XX, it has lead to deaths of 12 people.
A majority of artists and other creators believe that AI is "stealing" their work. There are also concerns about job losses due to AI.
Disclaimer: I have never experienced any parasocial relationship. My concerns of privacy, cognitive decline, hallunications, and sycophancy have stopped me from ever creating an account on AI sites, and hence from getting addicted to it. In past, I sometimes used Github Copilot for boring programming tasks. But now I have stopped that as well.
If you have finished reading this article, and still wish to continue then it’s your choice. I am not going to personally tell or force anyone to stop. I am also against the idea of goverment banning it. It does not bother me or anyone else.
If you want to stop then here are few things you can try. The final goal should be to completely stop relying on AI. First step would be to try to question yourself "why do I depend on AI?". When you reach a satisfactory answer, delete your pre-exisisting setup and start with a default one. Add a strict prompt to not discuss any other topic than academic or technical.