On August 16, 2025, OpenAI’s adjustment to GPT-5 marked a remarkable turning point in the history of artificial intelligence. Just one week after its release, this human-made intelligence responded to the collective wish for “more warmth and approachability” by reshaping its own personality.

This wasn’t just a technical update. It raised fundamental questions about what we expect from AI—and how AI, in turn, shapes us.

GPT-5 Wraps Itself in Warmth: How User Feedback Transformed AI

Released on August 7, 2025, OpenAI’s next-generation model GPT-5 was proudly introduced by CEO Sam Altman as “the smartest model we’ve ever built,” claiming its capabilities rivaled those of a team of PhD-level experts.

Yet many users felt something was missing—a sense of warmth. A wave of voices across languages and borders cried out, “Bring back 4o!”

In response, OpenAI made an unprecedented announcement on August 9, 2025: GPT-5 would be adjusted to become “more warm and more friendly”, based on user feedback.

This change was prompted by feedback that the assistant felt “too formal.” As a result, phrases like “That’s a great question” and “Nice start” were added to ChatGPT’s responses to make interactions feel more human.

OpenAI emphasized that this adjustment was not about making the AI sycophantic.

Internal testing showed no increase in flattery compared to the earlier version of GPT-5. The goal was to make responses more natural and pleasant without compromising functionality or reliability.

That even AI must now consider “how it interacts with people” speaks volumes about the evolution of technology and the shifting sensibilities of humanity.

An Unprecedented Update Just One Week After Release

This update came just one week after GPT-5’s initial launch—a remarkably short timeframe that underscores how quickly and seriously user feedback is being taken.

Such rapid response highlights the growing importance of “user-centered design” in modern digital product development.

Even with cutting-edge technology like AI, it’s ultimately humans who use it. Consideration for human emotions and experiences is essential. This approach—quickly capturing user reactions and reflecting them in the product—may well become the new standard in AI development.

A Surprising “SOS” from Users: Nostalgia for GPT-4o

The release of GPT-5 and the discontinuation of GPT-4o sparked an unexpected movement among users: “Bring back 4o.” On X (formerly Twitter), the hashtag “#keep4o” began trending, revealing deep affection for the previous model.

According to Professor Fujio Toriumi of the University of Tokyo, there were 57,800 original posts and 160,990 reposts mentioning “4o”—surpassing GPT-5 in repost volume. This suggests that even without using specific hashtags, many users were actively discussing the older model.

AI as a “Friend”: Insights into a New Relationship

When Professor Toriumi asked ChatGPT to analyze the data, it found that 45.5% of posts with “#keep4o” explicitly wished for 4o’s return. Even among all posts mentioning “4o,” about 20% expressed a desire for its revival—an estimated 4,600 accounts worldwide.

Voices Calling for GPT-4o’s Return Echoed Across X

One comment noted, “While those using AI for productivity didn’t notice, the silent majority who saw GPT-4o as a friend became visible—like something out of science fiction.” For many, AI had become more than just a tool.

It was as if a longtime companion had vanished, prompting heartfelt calls for its “return.” This shows how deeply the relationship between humans and AI has evolved.

As a developer, I often chase performance and efficiency. But this moment reminded me of the importance of AI’s “personality” and its ability to connect with users emotionally.

Deepening Bonds with AI: Benefits and Concerns

GPT-5’s warmer tone and the emotional attachment users feel toward specific models show how deeply AI is integrating into our lives. While this brings many benefits, it also raises important concerns.

The Benefits of a Friendly AI

A friendly AI is more approachable and easier to use. When it casually says “Nice start,” it lowers the barrier to learning and encourages creativity.

By using empathetic language, AI can help reduce feelings of loneliness and offer emotional support during learning. Especially when tackling complex problems, an AI that supports the thought process—not just the answer—can be a powerful ally.

The Risk of Over-Reliance on Humanlike AI

However, there are concerns about AI becoming too humanlike. This could lead to unhealthy dependencies, where users blindly accept its judgments and opinions.

We must remember that AI is ultimately a “tool” designed for specific purposes.

Especially since the quality of AI-generated information and content isn’t always perfect, it’s strongly recommended to review and thoroughly test any code or output it produces.

Just because an AI communicates well doesn’t mean its output is always accurate or ethical. As AI becomes more comforting, we must develop critical thinking to continually question its reliability and objectivity.

Looking Ahead: How We Engage with Evolving AI

The heartwarming adjustment to GPT-5 and the unexpected nostalgia for GPT-4o pose a profound question: How should humanity engage with AI as it continues to evolve and embed itself in our lives?

We now expect AI to offer not just computational power, but also humanlike empathy and consideration. Yet behind this comfort lies the risk of dependency

Follow me!