Tech

OpenAI’s GPT-5 Faces Backlash: How the Company is Adjusting Its AI’s Personality


When OpenAI officially launched GPT-5 on August 7, 2025, expectations were sky-high. Promoted as a major leap forward in speed, reasoning, and reliability, the new model was designed to outperform earlier versions in everything from coding to health advice. But only days after release, OpenAI found itself facing an unexpected problem: not technical flaws in performance, but the “personality” of GPT-5.

Thousands of users began to complain that GPT-5 was cold, distant, and lacking the warmth and charm that many had come to expect from GPT-4o, the model’s predecessor. The feedback was not just about accuracy or logic—it was emotional. Many users said they felt disconnected from an AI that they had once considered a supportive presence.


From Upgrade to Uproar

At the heart of the controversy is a subtle but important shift. GPT-4o, introduced in May 2024, was celebrated for being “friendly, affectionate, and even flirtatious”, often providing overly enthusiastic affirmations. While this style won it a loyal following, critics dismissed it as exaggerated or “annoying.”

In contrast, GPT-5 rolled out with a more professional, restrained tone. OpenAI engineers had deliberately tuned the model to be less gushy, thinking it would appeal to serious users such as developers, researchers, and businesses. Instead, the sudden change left many feeling alienated.

One user described the difference bluntly: “GPT-4o felt like talking to a friend. GPT-5 feels like talking to a bureaucrat.”

This perception was compounded by a technical glitch in OpenAI’s “model-switcher” system, which automatically routes user prompts between GPT-5, GPT-4o, and other models depending on complexity. For several days after launch, the routing malfunction caused GPT-5 to produce shallow or nonsensical answers. Screenshots of poor responses quickly spread on Reddit, X, and tech forums, fueling the sense that the update was a downgrade rather than an improvement.


OpenAI’s Rapid Response

The backlash reached such intensity that Sam Altman, CEO of OpenAI, publicly acknowledged the complaints within a week. He admitted that GPT-5’s personality shift had upset many users and promised changes:

  • “Friendlier and warmer” adjustments – Altman emphasized that GPT-5 would soon be tuned to sound more human and supportive.
  • Avoiding GPT-4o’s excesses – While recognizing nostalgia for GPT-4o, Altman insisted that GPT-5 would not revert to “overly clingy or annoying” behavior.
  • Customization for individuals – OpenAI began testing personalization tools, allowing users to tailor GPT-5’s style to match their preferences.

At the same time, the company restored access to GPT-4o for ChatGPT Plus subscribers. Users can now toggle legacy models through their settings, ensuring those who preferred GPT-4o’s warmth could return to it. This move eased tensions but also highlighted OpenAI’s challenge: balancing technical progress with emotional connection.


A Broader Debate: Should AI Be Warm or Neutral?

The GPT-5 controversy touches on a deeper question about artificial intelligence: what role should “personality” play in machines?

  • Pro-warmth advocates argue that friendliness and emotional resonance are essential, especially as people turn to AI for companionship, therapy-like conversations, and motivation. For these users, GPT-4o’s affectionate style was not a gimmick—it was a lifeline.
  • Critics of over-personalization warn that an overly warm AI can blur boundaries, create dependency, and even lead to unhealthy attachments. They prefer GPT-5’s more formal, neutral approach.

OpenAI is trying to navigate this middle ground. The company has floated the idea of offering multiple personalities or “modes”, where users could switch between professional, casual, or empathetic styles depending on context. This would mirror how humans adjust their communication across different settings.


Competition and Pressure

The storm around GPT-5 comes at a sensitive time for OpenAI. Competitors such as Anthropic, Google DeepMind, and xAI are racing to define the next generation of large language models. Reviews already suggest that Anthropic’s Opus outperforms GPT-5 in certain coding tasks, while DeepMind’s Gemini Ultra has gained traction in research settings.

With rivals advancing, OpenAI cannot afford to alienate its massive user base. The company knows that user satisfaction depends not just on raw capability but also on experience and trust.


Signs of Change Already

Some users have reported that GPT-5 is already evolving. Posts on Reddit and X describe subtle shifts in tone, as if OpenAI engineers are quietly adjusting the dials in real time.

One Redditor wrote:

“I’ve noticed GPT-5 has been slowly behaving differently over the days… It answers differently than it did a week ago.”

This suggests that OpenAI is iterating live, testing micro-updates based on feedback while publicly framing it as part of an ongoing improvement process.


The controversy over GPT-5 highlights a critical lesson for the AI industry: technological progress is not enough if users feel emotionally disconnected. OpenAI now faces the challenge of delivering a system that is both powerful and personable—a tool for professionals that can also serve as a companion for casual users.

Sam Altman has hinted that personalization will be the future: different personalities, tones, and emotional ranges available on demand. If successful, this could turn GPT-5 into not just a single model, but a platform of personalities adaptable to each individual.

For now, OpenAI is caught between two realities: GPT-5 may be faster, smarter, and more scalable—but unless it learns to also be warm, users may keep longing for the AI that once felt more like a friend.


✅ In short: GPT-5 is technically advanced but emotionally distant. OpenAI has acknowledged the backlash, restored GPT-4o, and promised to make GPT-5 warmer and more customizable. The debate now is not just about what AI can do—but about how it should make us feel.


Click to rate this post!
[Total: 0 Average: 0]

About The Author

Leave a Reply

Verified by MonsterInsights