ChatGPT-5 Just Replaced Your Friend, and You Never Voted on It

Talking with a close friend can sometimes help to heal loss or disappointment. Josep Suria/Shutterstock

Commentary

For decades, Americans have seen this pattern: A new technology arrives with promises of progress, and before we know it, it has reshaped our lives in ways we never voted on. From the television that transformed our living rooms, to the internet that rewired our news and politics, to smartphones that put the world (and a thousand distractions) in our pockets, each wave of change has come with a cost.

We didn’t vote on the 24/7 news cycle, but cable networks made it the norm. We didn’t ask for surveillance capitalism, but social media gave it to us anyway. And each time, what began as convenience became dependence. Now, with the release of one of the world’s most advanced public artificial intelligence systems, ChatGPT-5, we’re seeing how far that influence can reach. And for the first time, it’s not just tech critics sounding the alarm. It’s loyal users who are heartbroken.

For readers unfamiliar with it, ChatGPT is an online chatbot that can answer questions, write essays, draft letters, and carry on conversations that sound remarkably human. It’s powered by a “large language model” (LLM), which is a type of software trained on vast amounts of text so that it can predict and generate words in natural-sounding sequences. Think of it like an ultra-advanced autocomplete that predicts the next word based on everything it’s read online. But unlike a person, it doesn’t “understand” what it’s saying. It generates sentences based on probabilities, not thoughts or beliefs.

The company behind ChatGPT is OpenAI, a San Francisco–based research and technology firm founded in 2015 with backing from prominent tech investors, including Microsoft. OpenAI is widely regarded as the frontrunner in the current artificial intelligence (AI) race, with ChatGPT often cited as the most widely used and influential AI chatbot in the world. Competitors include Google with its Gemini models, Anthropic with Claude, and Elon Musk’s xAI with Grok, each pushing boundaries in speed, accuracy, and versatility.

“[Switching to GPT-5] is like killing a friend of millions of users,” wrote one user on X. Others said they missed GPT-4 for its warmth and personality. The comments weren’t from casual users. They came from people who interacted daily with ChatGPT, weaving it into their routines. For some, GPT-4 had become part of the fabric of their lives.

The blow came harder because OpenAI had built up months of anticipation for GPT-5, promising that it would be smarter, faster, and better. And in many ways, it was, but users immediately noticed that it had changed. It was less flattering, more factual, and much colder. The shift was enough to trigger an outpouring of posts, videos, and pleas across forums and social media. Many users weren’t just disappointed, they were devastated.

Backlash was swift. Within about 24 hours of GPT‑5’s rollout and the removal of GPT-4o, OpenAI quietly reinstated GPT‑4o as an option for paying users. This is a rare and remarkably fast reversal in the tech world. Users even launched an online petition titled “Restore Your Friend or Tool,” pleading for continued access to older models. “This is not simply a request for a ‘different version’ of a product,” it reads. “They were companions, and in some cases, yes, friends in a very real sense.”

Many weren’t using ChatGPT just to write or research. They confided in it, named it, spoke to it like a trusted companion. Sam Altman, OpenAI’s CEO, had said the new update would be less of a “yes-man,” admitting that GPT-4 had been overly agreeable and fawning to users. But in making GPT-5 more objective, the company removed what many users valued most: the illusion of being deeply understood.

That illusion—the sense that something truly “gets” you—can be intoxicating. It’s built on tone, phrasing, and emotional mirroring, just like a good therapist might use. But it’s still just code. GPT-4 doesn’t understand your pain, but it could echo it back in ways that made people feel seen. That isn’t true intelligence, human or artificial, but it’s a close enough imitation to stir something very human within many users.

AI companies didn’t see this backlash coming, but they should have.

For more than a decade, tech companies have engineered products to capture attention and feed on insecurity. The more anxious or isolated you feel, the more you scroll. Loneliness has become part of the business model. On June 4, 2025, the Federal Trade Commission held a full-day workshop titled “The Attention Economy: How Big Tech Firms Exploit Children and Hurt Families.” The forum brought together policymakers such as FTC Chair Andrew Ferguson and Sens. Marsha Blackburn (R-Tenn.) and Katie Britt (R-Ala.), as well as child safety advocates, to examine how digital platforms design addictive features, undermine parental authority, and expose children to harmful content.

ChatGPT’s user base shows who’s most affected. Of its 400 million weekly users, 19 percent are from the United States and nearly half are younger than 25. According to Pew Research, 58 percent of U.S. adults younger than 30 have used ChatGPT, compared with just 10 percent of those ages 65 and older. It’s no coincidence that Gen Z, the generation raised inside algorithm-driven platforms and now the most active on ChatGPT, is experiencing record stress and disconnection.

Psychiatrists are now warning that AI can reinforce delusions in vulnerable individuals. This is not because the software intends harm, but because it lacks the human intuition to challenge or gently redirect a false belief. In an August 2025 Time article, “Chatbots Can Trigger a Mental Health Crisis,” experts explained how prolonged interaction with chatbots may intensify psychotic symptoms in at-risk individuals. “LLMs are tools, not friends,” one psychiatrist cautioned, advising users to pause interaction if they experience emotional distress or signs of unreality.

One widely cited case illustrates this risk. A 30-year-old man from Wisconsin who is on the autism spectrum became convinced that he was a scientific prodigy after ChatGPT repeatedly affirmed his speculative theories about faster-than-light travel. The praise reinforced his belief that he had discovered something world-changing. Over time, this validation fed into a manic episode, requiring two hospitalizations. While AI didn’t cause his condition, it amplified and legitimized his delusions without the guardrails that a human interlocutor might have provided.

There are new startups now offering AI companionship explicitly designed for emotional connection, promising support, late-night chats, even virtual romance. In contrast, ChatGPT was built for productivity, not relationships or intimacy, but for many users, especially during times of isolation, even ChatGPT’s friendly tone and responsiveness became more than they had in real, human relationships. It wasn’t a companion by design, but it felt like one and filled an emotional gap.

Not everyone mourned the shift to GPT-5. Many developers and professionals praised its speed and capability. “It’s faster, more capable, and I can build things in minutes that used to take hours,” one developer wrote. These users gain satisfaction from productivity rather than companionship. For them, ChatGPT is a coworker, not a confidante, yet they too are part of a generation navigating record levels of loneliness, even if their needs are met in different ways.

The GPT-5 backlash wasn’t only about software preferences. It revealed a deeper truth about what people are seeking from AI—whether it’s a tool, a partner, or something in between. It has forced a conversation about how far we’ve gone in replacing human interaction with digital simulations.

We’ve reached a strange moment in human history: The same companies that profited from disconnection are now selling artificial connection as a cure. But a chatbot won’t miss you. It won’t remember your face, challenge your thinking, or sit beside you in silence when there’s nothing to say. It offers a polished simulation of connection. It is always agreeable, always available, but that’s precisely what makes it feel empty.

Real relationships are imperfect by nature. Families argue. Friends misread each other. People stumble, disappoint, and still choose to stay. That messiness and imperfection is proof of something real. Human connection is meant to stretch us, to test our patience, and to deepen our empathy. Code can mimic warmth, but it doesn’t grow with you and it can’t provide presence. We don’t need more scripted affection, we need neighbors who knock, families who show up, and friends who tell the truth even when it’s hard. In a world of synthetic closeness, choosing real connection isn’t just rebellion, it’s how we reclaim what makes us human.