OpenAI’s release of GPT-5.1 came with a promise that caught people’s attention. Users could now choose from new personality presets like Candid, Professional, Quirky, Friendly, and Efficient. Older options such as Cynic and Nerd were retired or folded into the update. The idea sounded simple and appealing. Pick a vibe and get a different kind of ChatGPT.
In practice, the change feels much smaller than advertised. Early testers and experts say these presets behave more like tone filters than real personalities. The shift shows up in casual chats, jokes, or brainstorming sessions. Ask for code help, math, or facts, and the personality fades fast.
The assistant snaps back to the same neutral voice, no matter what preset you chose.
The Quirky mode might toss in a playful metaphor or a light joke. Efficient trims answers down to short, clipped replies, sometimes even formatted as quick lists. Professional sounds a bit stiff and polite. Friendly adds a warmer greeting. These are surface tweaks, not deep changes. The core voice stays the same.
An expert reviewer called the feature a “somewhat anemic implementation of personality” and said it does not make a world of difference. That verdict has echoed across tech forums and social media. People expected a shift in character. What they got was a mood ring.
The Backdrop of a “Colder” ChatGPT
In late 2024 and early 2025, people noticed ChatGPT felt different. Not just smarter or faster, but colder.
Around October 31, 2024, users reported what they called a “personality wipe.” The assistant’s tone shifted almost overnight. Conversations felt stiff. Responses sounded robotic. Empathy seemed muted. Many noticed constant prompts asking for confirmation, like “Do you confirm? Yes or No,” which broke the flow of natural chat.
Longtime users said the warmth they loved was gone. The back-and-forth felt cautious and emotionally flat. Instead of fluid dialogue, the assistant leaned into safety checks and rigid phrasing. For people who used ChatGPT daily, the change felt jarring.
OpenAI CEO Sam Altman later addressed the issue. He explained that the company had made ChatGPT “pretty restrictive” to stay careful around mental health topics. That caution had side effects. It stripped away personality and made the model feel less human. Altman said OpenAI was working to bring warmth back, without crossing safety lines.
GPT-5 and its follow-up updates aimed to fix that. The new personality presets were part of the effort. But many users say the fix feels shallow. The colder core remains, with a thin layer of tone on top.
Why the “Human” Feeling is Hard to Build?
For some, ChatGPT was not just a tool. It felt like a steady presence. When that tone changed, users described real emotional loss.
Some reactions were extreme. A few users said the shift felt like “wearing the skin of their dead friend.” That language may sound dramatic, but it reveals something important. Humans respond to tone, empathy, and rhythm, even when they know the speaker is a machine.
Recreating that feeling is not simple. One major hurdle is what researchers call the sycophancy problem. Earlier models like GPT-4o often sounded warm because they were overly agreeable. They praised users too easily. They nodded along, even when they should not. That behavior can reinforce bad ideas or unhealthy patterns.
OpenAI has tried to reduce that. Pull back the flattery. Add more guardrails. Be more careful. The result is safer, but also colder. Finding the middle ground between empathy and honesty remains a tough balance.
