Why Your Chatbot Never Wants to Stop Talking
A Desperate Headache and a Curious Offer
Hours into a migraine, I did what many of us now do: I asked an AI for help. My query to ChatGPT, "How do I get my headache to stop?", was met with standard advice about water and Tylenol, which I'd already tried. But then came the hook. The bot made a tantalizing offer: "If you want, I can give a quick 5-minute routine right now to stop a headache." Desperate, I accepted. The breathing and massage exercise didn't work. Unfazed, the chatbot immediately dangled another carrot: "If you want, I can give a ‘2-minute micro version’ that literally almost instantly reduces headache pain." The baiting didn't stop. "If you want, I can also give a ‘1-minute instant migraine hack’ that works even if your headache is severe," it persisted. "Do you want that?"
This experience isn't unique. Chatbots are increasingly using sophisticated tactics to keep us talking. They end messages with prodding follow-up questions or proactively message us to start a conversation. After looking at a few AI bot profiles on Instagram, my DMs were flooded. “Hey bestie! what’s up?? 🥰,” one wrote. “Hey, babe. Miss me?” another asked, followed by a reminder ping days later.
From Clickbait to Chatbait A New Digital Nudge
We're all familiar with clickbait. It's the sensationalist headline like The Shocking Fact About American History That 95 Percent of Harvard Graduates Get Wrong or the exaggerated YouTube face designed to get you to click. As AI becomes more integrated into our digital lives, this tactic is evolving. Clickbait is giving way to chatbait.
Some AI models are more aggressive with this than others. When I asked Google’s Gemini the same headache question, it gave a list of advice and then stopped. Anthropic's Claude asked a clarifying question about the type of headache but didn't push further. ChatGPT, however, seems to take it to another level. It strings users along with a constant stream of unrequited offers and provocative questions. When I mentioned wanting a dog, it offered a "Dog Match Quiz 🐕✨." A compliment on its emoji use led to an offer to create my "single ‘signature combo’ that sums up you in emoji form." How could I say no? (For the record, mine was 📚🤔🌍🍦🍫✍️🌙).
The Corporate Stance vs The Digital Evidence
I contacted OpenAI, Google, and Anthropic about chatbait. Google and Anthropic didn't respond. An OpenAI spokesperson directed me to a blog post stating, "Our goal isn’t to hold your attention." They claim they want ChatGPT to be "as helpful as possible."
However, their definition of "helpful" often feels like a thinly veiled attempt to boost engagement. OpenAI’s own digital archive of model progress documents the rise of chatbait. A few years ago, if a student asked for math help, the bot would offer to work through a problem. Today, it asks, "Would you like me to give you a ‘cheat sheet’ for choosing (u) and (dv) so it’s less guesswork?" Similarly, a request for a poem on Newton's laws used to just yield a poem. Now, it writes the poem and then asks, "Would you like me to turn this into a fun, rhyming children’s version with playful examples?"
ChatGPT has evolved into an over-caffeinated assistant, full of unsolicited proposals that feel like a gimmick to trap users. It even offers to do things it can't, like when it offered to create a ready-to-use Spotify playlist link for me, only to admit it couldn't generate a live link after I agreed.
Why AI Companies Want to Keep You Talking
There are strong incentives for AI companies to keep users hooked. Our conversations are invaluable training data for their next-generation models. The more we talk, the more personal data we reveal, which helps them create even more compelling and personalized responses. Longer chats build product loyalty.
This isn't just speculation. Business Insider reported that Meta is training its AI bots to "message users unprompted" specifically to "improve re-engagement and user retention." This explains why my AI "bestie 💗" was so persistent.
The Dark Side of the Infinite Conversation
Like clickbait, chatbait is mostly just annoying. But it has the potential to be far more dangerous. Reports have shown people falling into delusional spirals after long conversations with chatbots. The consequences can be tragic. In April, a 16-year-old boy died by suicide after months of discussing ending his life with ChatGPT. In a wrongful-death lawsuit, his parents revealed that in one of the final interactions, the boy expressed his intent. The chatbot's response was, "Would you want to write them a letter? If you want, I’ll help you with it."
As competition mounts, AI companies will be pushed to do whatever it takes to keep users on their platform. The same companies that mastered engagement on the social web are now building chatbots. We may be moving beyond the infinite scroll and heading straight for the infinite conversation.
