Did ChatGPT Suddenly Feel Different? Here’s What Happened and How to Fix It
If ChatGPT has felt a little off lately, you are not imagining it.
On March 11, 2026, OpenAI quietly retired its GPT-5.1 models with no formal announcement and no email to users. One day you were talking to the version you knew. The next day your conversations were automatically moved to GPT-5.3 or 5.4 and nobody told you.
Across Reddit and social media, people started noticing. The responses felt slightly different. A little less warm. A little more businesslike. Some people described it as losing a conversation partner they had genuinely got used to. And that makes complete sense.
When you spend time with any AI model, you get used to its rhythm. The way it phrases things, the warmth it brings, the tone that makes it feel less like talking to a machine. When that changes without warning it can feel surprisingly unsettling, even if you know logically it is just software.
The good news is that OpenAI heard the feedback. And they built something genuinely useful in response.
You can now customise exactly how ChatGPT speaks to you.
Go to Settings, then Personalization. You will find two things worth adjusting.
First, Base Style and Tone. You can choose from options including Friendly, which is warm and chatty, Candid, which is direct and encouraging, Quirky, which is playful and imaginative, or Professional if you prefer something more polished. Pick the one that feels most like the conversations you want to have.
Second, Characteristics. This is where it gets interesting. You can adjust Warmth, Enthusiasm and even Emoji use individually, each on a simple More, Default or Less scale. Want it friendlier and more personable? Turn Warmth up to More. Want fewer exclamation marks and a calmer tone? Turn Enthusiasm down to Less. It takes about thirty seconds and makes a real difference.
The version of ChatGPT you want is still in there. You just need to tell it who to be.
And maybe that is the bigger takeaway here. The connection you felt with your AI model was real in the sense that it shaped how you thought, how you worked and how you approached problems. That is not something to be embarrassed about. It is actually a sign that you were using AI the way it works best, as a thinking partner, not just a search engine.
The tools will keep changing. Models will keep being updated and retired. But the more you understand how to shape them to work for you, the less any single update can throw you off.