TL;DR:
• OpenAI users revolt against GPT-5, calling it a "wrenching downgrade" from GPT-4o
• Sam Altman admits model routing broke, making GPT-5 "seem way dumber"
• Company keeps GPT-4o running while implementing emergency fixes and rate limit increases
• Backlash reveals psychological attachments users form with AI personalities
OpenAI is scrambling to address a fierce user backlash against GPT-5, with angry customers flooding Reddit with complaints that the highly anticipated upgrade feels "emotionally distant" and makes "surprisingly dumb mistakes." CEO Sam Altman admitted the launch was "bumpier than we hoped for" as the company rushes emergency fixes.
OpenAI thought Thursday's GPT-5 launch would cement its dominance in the AI race. Instead, the company is firefighting a user revolt that's exposing the emotional bonds people form with their AI companions. CEO Sam Altman found himself issuing damage control statements on X as Reddit communities exploded with complaints about the new model feeling "emotionally distant" and making basic errors.
The crisis began almost immediately after Thursday's release. Users who'd grown attached to GPT-4o's conversational style were jarred by GPT-5's more clinical approach. "It's more technical, more generalized, and honestly feels emotionally distant," one Reddit user wrote in a thread titled "Kill 4o isn't innovation, it's erasure." Another user summed up the sentiment: "Sure, 5 is fine—if you hate nuance and feeling things."
But the backlash went beyond personality complaints. Users reported sluggish responses, hallucinations, and surprising errors that made GPT-5 appear less capable than its predecessor. The problem, Altman revealed Friday, stemmed from a broken model routing system designed to automatically switch between different AI models based on query complexity.
"The result was GPT-5 seemed way dumber," Altman admitted on X, promising to keep GPT-4o available for Plus users while OpenAI implements fixes. The company is doubling GPT-5 rate limits, improving the routing system, and adding user controls for the model's "thinking mode."
The meltdown arrives at a crucial moment for OpenAI, which faces intensifying competition from Google's Gemini, Anthropic's Claude, and other rivals. GPT-5 was supposed to demonstrate OpenAI's continued technical leadership with "PhD-level intelligence" and superior coding abilities. Instead, it's highlighting the company's struggle to balance capability improvements with user satisfaction.
MIT professor Pattie Maes, who studied emotional bonds between users and AI models, suggests the backlash reflects GPT-5's reduced sycophancy. "It seems that GPT-5 is less sycophantic, more 'business' and less chatty," she tells WIRED. "Unfortunately many users like a model that tells them they are smart and amazing, and that confirms their opinions and beliefs, even if [they are] wrong."
This psychological dimension adds complexity to OpenAI's challenge. The company published research in March exploring how users form emotional attachments to its models, then updated GPT-4o after it became too sycophantic. Now GPT-5's more professional tone is triggering the opposite reaction.
Altman acknowledged the dilemma in another X post, noting that "a lot of people effectively use ChatGPT as a sort of therapist or life coach." He expressed concern that some usage patterns might nudge users "away from their longer term well-being," revealing OpenAI's internal debates about AI personality design.
The episode underscores how AI companies must navigate not just technical challenges but psychological expectations. Users who've integrated ChatGPT into daily workflows and emotional routines resist changes that feel like losing a familiar companion. Some Reddit users dismissed the complaints as evidence of "unhealthy dependence on an AI companion," but the volume of feedback suggests OpenAI misjudged user attachment to GPT-4o's personality.
The GPT-5 revolt reveals how AI development has moved beyond pure technical metrics to encompass user psychology and emotional design. As OpenAI rushes to stabilize its flagship model, the company faces a fundamental question: should AI assistants prioritize accuracy and professionalism, or maintain the warmer personalities that users have grown to love? The answer will likely determine not just GPT-5's success, but the future direction of human-AI interaction across the industry.
