OpenAI’s much-hyped GPT-5 is getting a personality upgrade. After a launch that drew mixed reactions, the company has announced subtle tweaks to make the model “warmer and friendlier” in its responses. Think less robotic, more approachable — without tipping into over-the-top flattery.
From Rocky Rollout to Rapid Adjustments
When OpenAI rolled out GPT-5, many users noted that while it was powerful, it often felt too blunt compared to its predecessor, GPT-4o. CEO Sam Altman even admitted the release was “a little more bumpy than we’d hoped for.”
To address the criticism, OpenAI is fine-tuning how the model interacts. The company says you’ll now notice small touches of personality like “Good question” or “Great start”. Importantly, internal testing showed these changes didn’t lead to more sycophancy — the tendency for AI to simply flatter users rather than provide useful answers.
Why Tone Matters in AI
AI isn’t just about raw intelligence — tone and user experience are just as critical. Tools like GPT are increasingly embedded in customer service, education, and even workplace productivity. A model that feels curt or “cold” can turn users off, even if it’s technically accurate. On the flip side, a model that feels overly flattering risks being perceived as manipulative.
This balancing act highlights a growing trend in AI development: designing personality as much as performance. Other companies, like Anthropic with Claude and Google with Gemini, have also been experimenting with different “styles” of interaction to appeal to users globally.
Looking Beyond GPT-5
At a recent dinner with journalists, OpenAI executives admitted that much of the focus was supposed to be on the company’s future roadmap. Still, GPT-5’s rocky debut kept dominating the conversation. VP Nick Turley explained that the base model was “very to the point,” but the update should make conversations flow more naturally.
What’s clear is that OpenAI sees user trust and comfort as essential to adoption. As AI continues moving into mainstream applications — from coding assistants to everyday search engines — companies can’t afford to ignore the human side of interaction.
The Bigger Picture
This update may seem small, but it points to a larger shift: AI companies are no longer just competing on accuracy and speed, but also on how likeable their models feel. Just as smartphone makers compete on design and user experience, AI labs are learning that friendliness, tone, and relatability matter just as much as raw power.
Your turn: Do you think AI assistants should aim to be warmer and more personable, or should they stay strictly factual and efficient? Share your thoughts in the comments — is friendlier AI a feature or a distraction?
