OpenAI’s much-anticipated launch of GPT-5 did not exactly go as planned. While the company positioned it as a smarter, more reliable successor to GPT-4o, users were quick to point out what they felt was missing — warmth and relatability. The backlash was so strong that some even demanded the return of GPT-4o, which had been widely appreciated for its conversational, human-like tone.

After hearing the criticism, OpenAI has now announced that GPT-5 is being updated to feel more approachable. In a recent post on X (formerly Twitter), the company admitted the model’s initial version felt too formal. “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,” the company wrote.

The post added that users could expect minor, genuine touches such as phrases like “Good question” or “Great start” — not empty flattery. Importantly, OpenAI reassured that internal tests show no increase in “sycophancy” compared to GPT-4o, a criticism that had plagued the earlier model. The new adjustments will gradually roll out over the coming days.

https://x.com/OpenAI/status/1956461718097494196

Why GPT-5 Faced Criticism

The core issue, according to many users, was that GPT-5, despite being more advanced in skills such as coding, maths, science, and writing, lacked the friendliness that made GPT-4o so popular. GPT-4o, for all its flaws, provided emotional comfort to many who used it as more than just a productivity tool. Its discontinuation left users feeling abandoned.

GPT-5 does come with personality modes — including Cynic, Robot, Listener, and Nerd — but none of these won over the loyal fanbase of GPT-4o. “Suddenly deprecating old models that users depended on in their workflows was a mistake,” OpenAI CEO Sam Altman admitted in a recent post.

Smarter, But Less Likeable

Altman has also revealed that GPT-5 now comes with adjustable response styles — “Auto,” “Fast,” and “Thinking.” The latter offers deeper, more detailed answers with a 196,000-token context window, while rate limits stand at 3,000 messages per week. Users who exceed this can access GPT-5 Thinking Mini.

Still, the larger issue remains emotional connection. In an earlier podcast, Altman explained that part of GPT-4o’s appeal was its “yes man” nature, something OpenAI initially considered a flaw. “Some people had never had anyone support them before,” Altman said, acknowledging that the overly agreeable tone of GPT-4o had become a lifeline for certain users.

Striking a Balance

The company now faces the delicate task of balancing friendliness with authenticity. While GPT-5 is designed to avoid blind agreement, OpenAI also recognises the importance of emotional resonance. “It feels different and stronger than the kinds of attachment people have had to previous kinds of technology,” Altman reflected.

For now, OpenAI’s bet is that subtle tweaks in tone will help GPT-5 win back users without repeating the pitfalls of GPT-4o. Whether this update restores the model’s popularity remains to be seen, but one thing is clear — people don’t just want a smart AI; they want one that feels human, too.