OpenAI Softens GPT-5 After User Pushback

When OpenAI launched GPT-5, many expected a smooth leap forward.

Instead, the release sparked mixed reactions. Some users said the new model felt too stiff compared to GPT-4o, which had a more approachable style.

Even OpenAI’s CEO, Sam Altman, admitted the rollout was “a little more bumpy than we’d hoped for.”

A Warmer, Friendlier AI

To address concerns, OpenAI has rolled out an update designed to give GPT-5 a more natural, human touch.

Instead of sounding overly formal, the model now adds subtle phrases like “Good question” or “Great start.”

The company says these aren’t meant as empty compliments but as genuine, conversational cues.

What’s changing?

• Friendlier tone
  
• More conversational feedback
  
• Still avoids over-the-top flattery

Internal testing shows the adjustments don’t make GPT-5 more sycophantic than before, which was one of the concerns users had early on.

Why the Change Matters

Think about it: nobody enjoys talking to someone who always sounds like a robot. People naturally connect better when conversations feel warm.

That’s true whether it’s a teacher encouraging a student or a co-worker giving feedback. OpenAI seems to be applying the same idea here.

Before Update	After Update
Direct, no extra warmth	Adds encouraging phrases
Felt blunt to some users	More approachable tone
Focused only on answers	Balances accuracy with friendliness

Looking Beyond GPT-5

At a recent dinner with journalists, OpenAI leaders tried to steer conversations toward the future of AI. Still, most of the attention circled back to the GPT-5 launch.

According to VP Nick Turley, the model was originally “very to the point.” Now, the shift aims to make interactions less mechanical and more relatable.

The Bigger Picture

This update isn’t just about making an AI sound nicer. It’s about trust.

If people feel comfortable with a tool, they’re more likely to use it in their daily lives, whether that’s for school, work, or creative projects.

So the real question is: will this warmer GPT-5 win back those who preferred GPT-4o?

Time will tell, but OpenAI seems determined to bridge the gap.