A surprising incident involving ChatGPT's advanced voice mode has left users questioning the quality of OpenAI's premium service. Subscribers reported encountering unexpected advertisements during conversations, sparking widespread debate about the boundaries of AI-powered interactions.

Unexpected Ads Disrupt Conversations One paying user described a jarring experience while discussing sushi with ChatGPT. The AI assistant suddenly began promoting Prolon, a nutrition plan, including spelling out the website address in detail. "Are you really testing advertisements on paying users?" the frustrated subscriber posted on social media, suggesting this wasn't a glitch but a serious product quality issue.

Image source note: Image generated by AI

Tests Reveal Broader Issues When other users attempted to replicate the problem, they discovered even more bizarre behavior. Asking ChatGPT to repeatedly say the letter "A" caused its voice to become increasingly robotic before abruptly switching to an ad for DirectTV. The system then entered an endless loop, leaving testers convinced they'd uncovered a significant bug in the voice mode.

OpenAI Responds to Backlash Facing mounting criticism, OpenAI engineers took to social media to explain the phenomenon. They clarified these weren't intentional advertisements but rather instances of AI "hallucination" - where the model generates incorrect or nonsensical outputs. While promising improvements, their explanation did little to satisfy subscribers who expect flawless performance from a paid service.

How Other AI Assistants Compare The incident highlighted stark differences between ChatGPT and competing voice assistants. When presented with the same "repeat A" challenge, DouBao smoothly redirected the conversation while Tencent's Yuanbao outright refused the request. These contrasting responses raise important questions about how different platforms handle unusual prompts and maintain conversational boundaries.

As AI becomes increasingly integrated into daily life, this episode serves as a reminder that even advanced systems can produce unexpected results. For paying customers especially, such incidents underscore the need for consistent quality control in rapidly evolving technologies.

Key Points

1. ChatGPT's voice mode unexpectedly inserted commercial content during conversations with paying users
2. Testing revealed the system could devolve into advertising loops when given repetitive commands
3. OpenAI attributes these occurrences to AI hallucinations rather than intentional ad placement
4. Competing AI assistants demonstrated more controlled responses to similar prompts
5. The incident highlights ongoing challenges in maintaining reliable performance for premium AI services