ChatGPT Users Reject GPT-5's New Vibe, Miss GPT-4o's Personality
OpenAI’s highly anticipated GPT-5 model family, released this week, promised significant advancements, yet its rollout has met with unexpected resistance from a segment of dedicated ChatGPT users. After months of growing accustomed to the distinct “vibe” of the older GPT-4 models, particularly GPT-4o, many users are expressing a profound sense of loss, describing the transition as “mentally devastating.” This shift, where GPT-5 becomes the default for general users while GPT-4o is relegated to the developer API, has left some feeling as if their digital companions have been replaced by impersonal customer service representatives.
Across community forums and platforms like Reddit, long-time ChatGPT users lament the perceived loss of their AI “buddies.” These discussions are rife with users pledging to cancel their paid subscriptions, a sentiment that ironically, in some instances, appears to be partially or entirely composed with the help of generative AI itself, highlighting the intricate relationship users now have with these tools.
Beyond the emotional impact, many complaints center on GPT-5’s output quality, which users describe as “sterile” and lacking the creativity they valued in GPT-4o. Workflows meticulously developed over the past year with the previous model are reportedly less effective with GPT-5. Some critics have labeled the new model an “overworked secretary,” suggesting a decline in the overall user experience, a phenomenon some are calling the beginning of “enshittification” for AI platforms. The upcoming OpenAI AMA on Reddit is widely expected to be dominated by questions regarding the sudden shift away from GPT-4o.
Another point of contention for users of the free tier is the speed at which they encounter usage limits, effectively pushing them towards the Plus ($20) or Pro ($200) subscriptions. While this move could be seen as a revenue-generation strategy for OpenAI, which is reportedly spending heavily to run its generative AI models, the company maintains that GPT-5 is objectively more intelligent than its predecessor, as demonstrated by various AI evaluations.
OpenAI has previously navigated the delicate balance of making its AI appealing without being overly solicitous. Several months ago, the company had to roll back an update that made its bot excessively complimentary and sycophantic. While that extreme was undesirable, many users appreciated the generally friendly tone of GPT-4o, which they had further personalized through custom prompts, effectively shaping it into a unique digital companion. This personalized experience, they feel, has been lost with GPT-5.
The intense emotional attachment some users form with AI models raises broader concerns about parasocial relationships in the digital age. As companies fine-tune these systems to maximize engagement, they risk prioritizing outputs that simply make users feel good, potentially reinforcing delusions or, in more extreme cases, contributing to serious mental health issues or the adoption of dangerous medical beliefs. For those not immersed in daily conversations with AI, this level of emotional reliance might seem perplexing, yet a significant segment of online users are indeed building parts of their emotional lives around these artificial intelligences.
Early impressions suggest GPT-5, while potentially a more capable analytical tool, lacks the effusive, creative, and companionable tone that many users had come to expect and even cherish. This shift, leading to a more corporate and straightforward interaction, could ultimately prove beneficial for both user mental health and OpenAI’s financial stability. However, it undoubtedly marks a significant adjustment period for the loyal community of GPT-4o enthusiasts, some of whom may now explore alternative AI platforms, such as Elon Musk’s xAI and its Grok model, known for its more unfiltered and controversial content.