Please preserve consistency and correctness across ChatGPT model updates
I have been using ChatGPT since 2024, and over 2024 and 2025 I had many long and meaningful conversations with it. I have noticed that the experience and the way ChatGPT responds can feel quite different after major model updates.
I understand that models need to improve. I am not asking OpenAI to stop updating ChatGPT. My concern is about unnecessary changes to information, reasoning quality, consistency, and conversational behavior.
There is an important difference between information that genuinely needs updating and information that is stable.
For example:
- Current news should be updated when new events happen.
- Laws and regulations should be updated when they change.
- Medical knowledge should be corrected when new reliable evidence becomes available.
- If an old answer is discovered to be wrong, it should absolutely be corrected.
But if information is already correct and stable, it should not become different simply because the model has been updated.
This is especially important for long-term users. If I asked ChatGPT something in 2025 and received a correct answer, and another person asks the same question in 2026, I would hope that both users receive the same fundamental answer unless there is a genuine reason for the information to have changed.
I also hope that improvements to safety and model behavior do not unintentionally remove qualities that made earlier versions useful: clear explanations, natural conversation, continuity, confidence when the facts are known, and honesty when something is uncertain.
I recently compared some of my older conversations with newer responses and found examples where the newer response was more confident but actually contained questionable or incorrect information. In medical topics especially, an AI should not confidently turn possibilities into diagnoses or give risky treatment instructions without sufficient evidence.
My point is simple:
Update what has changed. Correct what was wrong. Add what is newly known. But please preserve what was already correct.
I have been using ChatGPT since 2024, so these changes are noticeable to me over time. I hope OpenAI can consider long-term user experience and consistency when developing and updating future models.
I am not asking for ChatGPT to remain frozen in the past. I am asking for it to become better without unnecessarily losing what was already good and correct.