ChatGPT’s Model Reversal: How User Attachment Forced OpenAI to Rethink Its GPT-5 Strategy

When OpenAI rolled out its highly anticipated GPT-5 model last week, the company likely expected applause for its advancements in AI capability. Instead, what followed was an unexpected wave of user backlash that forced the tech giant to quickly reverse course on a key decision: the retirement of older models, particularly the beloved GPT-4o. This episode reveals a fascinating dimension of human-AI interaction that even OpenAI didn’t fully anticipate – users don’t just use AI models; they form attachments to them.

“In retrospect, not continuing to offer 4o, at least in the interim, was a miss,” admitted Nick Turley, OpenAI’s head of ChatGPT, in a recent interview with The Verge. What surprised OpenAI was the “level of attachment” people had developed to GPT-4o. As Turley explained, “It’s not just change that is difficult for folks, it’s also the fact that people can have such a strong feeling about the personality of a model.”

The initial plan for GPT-5 was straightforward and seemingly logical: create a unified system with a smart, efficient model for most queries, a deeper reasoning capability for complex problems, and a real-time router that would automatically decide which approach to use based on the context. This design would eliminate the need for users to navigate a complex model picker – something OpenAI CEO Sam Altman has publicly expressed frustration with. The company believed this simplification would benefit the vast majority of ChatGPT’s 700 million weekly users who only use the default model.

However, what OpenAI didn’t fully account for was how personalized the AI experience had become for many users, particularly power users who pay for ChatGPT Plus or Pro subscriptions. When GPT-4o disappeared overnight, the reaction was swift and emotional. Users took to social media platforms like Reddit and X (formerly Twitter) to express not just dissatisfaction but what could only be described as grief. Many lamented the loss of GPT-4o’s particular “personality” and expressed frustration at having workflows disrupted without warning.

Responding to the backlash, OpenAI quickly reintroduced GPT-4o as a default option for paying users. Altman announced via X that “4o is back in the model picker for all paid users by default. If we ever do deprecate it, we will give plenty of notice.” Additionally, paid users now have access to a “Show additional models” toggle in ChatGPT web settings, which adds legacy models like o3, GPT-4.1, and others.

Rather than sticking with the simplified approach, OpenAI has now expanded GPT-5’s options to include multiple modes: Auto, Fast, Thinking, and Thinking mini. The Auto setting works as originally intended, letting ChatGPT determine how to respond, while the other modes give users more specific control over the model’s behavior. As Altman explained, “Most users will want Auto, but the additional controls will be useful for some people.”

The irony is palpable: a model launched to simplify user choices now offers more options than ever. TechCrunch aptly noted that “ChatGPT’s model picker now seems to be as complicated as ever, suggesting that GPT-5’s model router has not universally satisfied users as the company hoped.”

The technical challenges didn’t help either. GPT-5’s model router reportedly faced issues on launch day, which caused some users to perceive the new model as less capable than its predecessors. This combination of technical hiccups and the emotional response to losing familiar models created a perfect storm of criticism.

OpenAI’s swift response to the feedback demonstrates the company’s awareness of how crucial user satisfaction is to its business model. With ChatGPT’s mobile app alone generating an astonishing $2 billion in revenue since its launch in May 2023, according to recent analysis by Appfigures, maintaining user loyalty is paramount. The app has earned $1.35 billion in 2025 alone, representing a massive 673% increase over the same period last year. With that kind of growth at stake, addressing user concerns quickly was a business imperative.

Beyond the immediate changes, Altman hinted at a longer-term solution: “We are working on an update to GPT-5’s personality which should feel warmer than the current personality but not as annoying (to most users) as GPT-4o,” he wrote. “However, one learning for us from the past few days is we really just need to get to a world with more per-user customization of model personality.”

This acknowledgment represents a significant shift in thinking about AI development. Rather than assuming all users want the same experience from an AI assistant, OpenAI is recognizing that different users may prefer different AI “personalities” – some might want a more conversational, warm assistant, while others might prefer a more straightforward, facts-only approach.

The episode also highlights an important aspect of human psychology that AI companies are now confronting: the emotional connections people form with technology. Similar phenomena have been observed elsewhere in the AI space, such as when hundreds of people in San Francisco held a funeral for Anthropic’s AI model Claude 3 Sonnet when it was taken offline.

For OpenAI, the GPT-5 rollout has provided valuable lessons about managing user expectations and transitions between models. While the company remains committed to advancing AI capabilities with models like GPT-5, it now recognizes the need to better understand and accommodate the attachments users form with specific AI personalities.

As AI becomes more integrated into daily life, companies will need to consider not just the technical capabilities of their models but also the emotional responses they evoke. The line between tool and companion continues to blur, creating new challenges for AI developers who must now consider the human side of human-AI interaction more carefully than ever before.

In the meantime, ChatGPT users can take comfort in knowing that their favorite models won’t disappear without warning, and that OpenAI is working toward a future where AI personalities can be customized to individual preferences. This evolution in thinking could lead to more personalized AI experiences that better serve diverse user needs while maintaining the technical advancements that push the field forward.

Related Products

For those interested in exploring AI technology further, check out these related products:

• AI: The Great Disruption: How Artificial Intelligence Will Reshape Jobs, Society, and the Future of Humanity – A comprehensive book on how AI is changing our world
• Nothing Ear Wireless Earbuds with ChatGPT Integration – Experience ChatGPT integration with these wireless earbuds
• AI Voice Recorder with App Control – Advanced AI technology for transcription & summarization
• 64GB Voice Recorder with AI Noise Reduction – Great for recording lectures, meetings, and interviews
• Mastering Flask Web and API Development – Learn to build and deploy apps across web, APIs, and mobile platforms