What’s wrong with GPT? This app has REALLY gone down in quality.
Viral complaints claim GPT-4o now gives flagged, unusable answers on simple purchase queries.
A viral Reddit post has ignited a significant discussion about a perceived quality regression in OpenAI's flagship GPT-4o model. Users are reporting that the AI assistant has become 'straight up unusable' for practical tasks, particularly those involving product recommendations and comparisons. The original poster detailed experiences where simple queries about purchasing a Jeep Grand Cherokee or comparing trucks resulted in the AI flagging the conversation and even recommending Dialectical Behavioral Therapy, indicating a severe over-application of safety filters. This suggests a recent backend update may have shifted the model's behavior to prioritize extreme caution over helpfulness, breaking core functionality for a segment of users who rely on it for consumer research.
The technical implications point to a potential conflict between OpenAI's safety alignment efforts and the model's utility for everyday tasks. The complaints are specific: the model fails to give straight answers on topics like 'how to make pierogis' or comparing city data, domains not typically associated with high risk. This indicates the safety filtering may be overly broad or mis-calibrated, affecting neutral commercial queries. For professionals and consumers who integrated GPT-4o into workflows for product research and data comparison, this regression creates a reliability gap. The situation highlights the ongoing challenge for AI developers in balancing safety guardrails with practical utility, and whether this is a temporary bug or a deliberate but poorly implemented policy shift will be crucial for user trust moving forward.
- Users report GPT-4o now flags harmless product queries (e.g., Jeep purchase) and suggests therapy.
- The model's perceived over-caution makes it unreliable for comparing vehicles, electronics, and city data.
- The issue suggests a recent safety update may have degraded practical utility for consumer research tasks.
Why It Matters
If AI assistants become unusable for basic product research, it undermines their core value proposition for millions of users.