A disruptive issue has recently emerged in the ChatGPT Web UI related to irrelevant prompt suggestions. During active development, entirely unrelated prompts—such as “Draw a duck” or “Suggest a delicious restaurant”—occasionally appear, occupying nearly half of the input area. Worse still, attempting to ignore these suggestions and moving the cursor upward unintentionally selects one of them, overwriting the ongoing input. Thankfully, the original input can be restored using Ctrl+Z.
Given its inconsistent reproducibility, I believe this behavior is unintended, possibly resulting from an experimental feature mistakenly activated in my production environment. I trust the OpenAI development team and am confident such a clearly problematic UI decision would not be intentionally implemented.
Ideally, prompt suggestions should directly relate to the current thread, predicting likely next user inputs (similar to Perplexity’s implementation). Furthermore, once the user starts typing their own input, any suggestions should disappear immediately.
Request:
Please treat this as an urgent usability issue requiring prompt attention and correction.