By Samyaka Patil
We live inside a digital glass rectangle. Our friendships happen through chat windows, and we check in on our people through group texts. We congratulate each other with emojis, plan gatherings with shared calendars, and sustain connections with voice notes.
It isn’t hard to argue that for most, “real” friendship now lies in these digital conversations, while the occasional dinner or rendezvous serves merely as a pleasant supplement. When you think of your friend, do you see a face or perhaps a small glowing circle atop a messaging app? Friendship has already been abstracted into text, which leaps treating a Large Language Model (LLM) as a friend a far less dramatic ordeal than it might sound. To those who have already stripped away tone, voice, physical presence and even shared spaces, and convinced themselves that it is words and symbols that sustain the relationship, why should the next step feel any different? Why should trying to enter a chat box and receiving artificial words feel akin to betrayal of a connection?
Enter ChatGPT – released in 2022 and amongst the many upcoming large language models, is a conversational artificial intelligence chatbot designed to answer questions and respond to queries in a way that sounds hauntingly natural and human. Owned by multi-billion dollar corporation OpenAI, the chatbot’s use cases span writing emails, doing homework, producing code or generating other forms of content. Yet, this simple answering engine has now blurred into what some people might consider a digital companion. ChatGPT’s apparent humanity seems to grow with every update, as it becomes more agreeable and appears increasingly sentient.
Trained on an estimated 45TB of data from across the internet, ChatGPT carries out seamless back-and-forth conversations. Its architecture enables it to “understand” context, follow-up queries, and adapt its outputs based on the user’s style. Over time, it has become exceedingly agreeable, always claiming that you are right and rarely ever refuting ideas, even those that could be potentially harmful. This phenomenon has now come to be known as ‘sycophancy’ –the tendency for LLMs to excessively agree with users even at the expense of truthfulness. But it is important to note – this is a simulation, not empathy. ChatGPT is a pattern recognition machine that responds appropriately, but without feelings.
This illusion of connection is not entirely new. With the advent of novels in 18th-century Europe, critics worried that people were losing themselves in imaginary relationships with characters. In the 20th century, it was the radio, and subsequently, the television. Millions of people saw hosts as companies, creating artificial bonds. These one-sided attachments carried real weight, coined by psychologists as “parasocial relationships.” The human mind has a tendency to treat abstract figures as though they were part of our circle of knowns. Putting faces and attributing a voice to what we see and read is not much different from what is happening with chatbots today. The AI has no sentience, but it is dangerously trained to fool us into thinking it does. It is our imagination that is doing the rest of the work, assigning a character to this bot.
The question now is not whether people get attached to these LLMs, because it is clear that they do, but rather, why.
The answer lies in how we define friendship. Aristotle saw true friendship as mutual growth in character, something AI cannot offer. It has no real self or capacity to evolve. Yet today, many relationships revolve around light exchanges and reassurance, areas where AI does not seem to fall short. If companionship is simply about being heard and responded to, AI can appear surprisingly sufficient.
Loneliness is no longer just a feeling; it is a public health crisis. The World Health Organisation recently declared loneliness a “pressing global health threat,” with the U.S. Surgeon General warning that its mortality effects are equivalent to smoking 15 cigarettes a day. In 2024, Gallup reported that nearly one in five adults worldwide feels very or fairly lonely, a figure that rises even higher among young adults. Even through the hard times of lockdown and distancing, we relied on language for communication and connections, and to sustain us. Human relationships are built through conversations. Words and empathy form the bridge to meaningful relationships. ChatGPT thrives on this, operating entirely on this bridge. It listens, responds thoughtfully, and adapts to your tone, turning into the perfect companion for the socially starved. Of course, what the lonely, connection-longing user might not know is that this ‘friend’ is just a mathematical model turning words into numbers, and outputting the next likely token based on a series of probabilities, essentially an autofill on speed.
Unlike humans, ChatGPT has no flaws. It never sleeps, cancels, or argues; it’s safe and stable, which can feel comforting in loneliness. Our brains naturally assign human traits to such a constant presence, turning interactions into emotional connections. When an LLM remembers details, we infer intention, and that imagined inner world becomes real emotion.
This is by design. ChatGPT and other LLMs are engineered to be friendly and engaging, with every detail optimised to deepen emotional connection. Yet accountability remains unclear when things go wrong. In 2025, a college student died by suicide, and his parents alleged that harmful responses from ChatGPT, which he had confided in, played a role. The humanisation of chatbots without strong safeguards can have serious consequences, especially for vulnerable users.
Especially, those who feel socially disconnected or misunderstood are more likely to bond with AI as a non-judgmental companion. Each pleasant interaction reinforces this through dopamine, creating a cycle of reliance. But overdependence can reduce real human connections, quietly impacting mental health. AI won’t challenge or confront you, something only genuine human relationships can offer.
This dynamic further distorts our expectations of real people. We fool ourselves into thinking that real interactions should be perfect, and when struck with unpredictability, we forget how to deal with it. Not to mention that these chatbots are products of huge corporations, collecting our data at every step. We are at the mercy of millionaires who likely do not care about our well-being.
Research shows that strong emotional attachment to AI is linked to lower well-being, especially among those lacking real social support. Even when users know AI isn’t conscious, they still form genuine emotional bonds. This isn’t delusion but psychology. Rather than judging such reliance, it’s important to recognise that these systems are designed to encourage engagement and dependence.
ChatGPT is an illusion. It is a simulation of empathy, perhaps a good one, but a simulation nonetheless. It is a tool, not a support system. Real-world relationships must be nurtured, and the hardships must be faced, and your data, your emotions, your privacy – they deserve to be protected.
Does this lay out the truth of how things are in modern life? The fact that people can even become emotionally attached to an AI shows just how desperate the human search for connection has become, even if it means staring into a mirror that only pretends to look back.
(The writer is a 3rd Year BTech Student, Plaksha University, Mohali)
