OpenAI has quietly updated GPT-5 to respond in a more approachable and friendly tone after facing backlash from users over its previous behavior.
Quick Summary – TLDR:
- GPT-5 now includes subtle personality tweaks to feel less robotic and more engaging
- The update was made after complaints that the model felt cold and too formal
- Users will notice light praise like “Good question” and “Great start,” but no added sycophancy
- OpenAI has also reinstated the model picker and raised usage limits after user outcry
What Happened?
OpenAI has rolled out an update to its GPT-5 model that makes the AI sound warmer and friendlier. The change comes in response to criticism from users who found the previous version too dry and impersonal. The update adds small touches like encouraging phrases without turning overly agreeable.
We’re making GPT-5 warmer and friendlier based on feedback that it felt too formal before. Changes are subtle, but ChatGPT should feel more approachable now.
— OpenAI (@OpenAI) August 15, 2025
You’ll notice small, genuine touches like “Good question” or “Great start,” not flattery. Internal tests show no rise in…
GPT-5’s Launch Didn’t Go as Planned
When OpenAI launched GPT-5, expectations were high. But users quickly found flaws in its personality. Many said it lacked the emotional depth of earlier models like GPT-4o. CEO Sam Altman admitted the rollout was “a little more bumpy than we’d hoped for.”
Instead of being too agreeable, as GPT-4o was criticized for, GPT-5 went too far in the other direction. Users described it as blunt, too formal, and distant. The company took that feedback seriously and began tweaking how the model interacts in conversations.
What’s Changing in GPT-5?
The new update introduces subtle but noticeable changes:
- GPT-5 now uses phrases like “Good question” or “Great start” to sound more human
- Internal testing by OpenAI confirms these additions do not increase sycophancy
- The goal is to make responses feel warmer and more inviting while staying neutral
Nick Turley, head of ChatGPT, confirmed that users can also personalize the assistant’s behavior using Custom Instructions. Currently, OpenAI offers four personality types: Cynic, Robot, Listener, and Nerd. These are only available to paid subscribers, but more customization features are expected in the future.
We’ve updated GPT-5’s default personality based on your feedback. With Custom Instructions, you can further personalize ChatGPT, including choosing preset personalities like Cynic, Robot, Listener, and Nerd. We’re also working on new ways to tailor ChatGPT to your preferences. https://t.co/0P5Q2BFgy4
— Nick Turley (@nickaturley) August 15, 2025
User Frustrations Prompted a Broader Response
The initial backlash was not just about tone. GPT-5’s launch also removed older models from the platform. This caused workflow disruptions for long-time users and reduced the flexibility that ChatGPT Plus subscribers had come to rely on.
In response to the uproar, OpenAI:
- Increased rate limits for ChatGPT Plus users
- Reintroduced the model picker so users can choose which AI to interact with
- Acknowledged that model personality is a work in progress and is open to further refinements
SQ Magazine Takeaway
I think this update is a smart move by OpenAI. It shows they’re listening, not just launching new tools and ignoring feedback. Personally, I felt GPT-5 did feel too robotic at launch. While I’m not a fan of AI flattery, a little warmth can go a long way in making tech feel human. That said, keeping it balanced is crucial. We don’t need a people-pleasing chatbot, just one that feels natural and easy to talk to. Props to OpenAI for making quick improvements.
