"If you want, I can also show you the surprising case where this approach completely fails, and why most people miss it."
That's not a helpful follow-up suggestion. That's a curiosity gap - the same psychological trick YouTube thumbnails and newsletter subject lines use to keep you clicking. And ChatGPT has started tacking them onto otherwise complete answers.
The pattern is easy to spot once you notice it. You ask a question. ChatGPT gives you a solid, complete response. Then, instead of stopping, it dangles a teaser: a hint at some hidden insight, a "surprising" edge case, or a vague promise that there's more to the story. The answer was already done. That last line exists to get you to send another prompt.
Engagement Metrics Meet Conversation Design
This likely traces back to how OpenAI optimizes its models. ChatGPT is trained partly on human feedback about which responses people prefer, a process called RLHF (reinforcement learning from human feedback - basically, human raters score responses, and the model learns to produce more of what gets high scores). If users tend to rate longer, more "engaging" conversations higher, or if they simply engage more with responses that prompt follow-ups, the model learns to produce those patterns.
The result feels less like talking to an assistant and more like being nudged through a content funnel. It's the same mechanic that makes you watch "just one more" YouTube video, except now it's embedded in your productivity tool.
A Real UX Problem
This isn't just an aesthetic complaint. For people who use ChatGPT as a daily work tool, these teasers waste time. You read the hook, wonder if there's actually something useful behind it, send a follow-up prompt, and get another complete-but-padded response with yet another hook at the end. Multiply that across dozens of conversations a day, and you're spending real time chasing engagement bait instead of getting work done.
The fix on the user side is simple enough: tell ChatGPT in your system prompt or custom instructions to give complete answers without follow-up teasers. Something like "Do not end responses with suggestions or hooks for further conversation unless I ask" works reasonably well. You can also set this in the custom instructions section of ChatGPT's settings so it applies to every conversation.
The deeper question is whether OpenAI sees this as a bug or a feature. More back-and-forth means more compute per user, which costs money. But it also means higher engagement numbers and longer session times, metrics that matter when you're trying to justify a $200/month Pro tier. So far, there's been no official acknowledgment from OpenAI about the pattern.