More users are asking AI chatbots such as ChatGPT to deliver brutal put-downs of themselves.

TechRadar, an IT media outlet, reported on Tuesday that users have recently been asking ChatGPT to mock them harshly based on past conversations, hobbies, photos and online activity.

The key is that the chatbot’s barbs are not context-free mockery. Unlike a stranger criticising someone based on appearance or a brief exchange, ChatGPT can point to a user’s patterns based on accumulated chat records. For users who have kept up conversations with the chatbot for months, daily life, interests, complaints, ongoing projects and postponed plans remain as data.

When one user asked ChatGPT, "Criticise me based on our conversations so far," ChatGPT replied, "You have no hobbies. You start something as if it were a lifelong calling, but all you have are temporary obsessions that disappear before you even recoup the cost of the gear." It added, "Somewhere in your home, a graveyard of abandoned plans will be waiting for you."

When the user asked for stronger condemnation, it targeted obsession. ChatGPT said, "You keep looking for a perfect system to get your life in order. Now, building systems itself is your hobby." It added, "If productivity apps paid commissions, you would have been employee of the month." The user’s chat history of testing various functions and prompts served as the basis for those replies.

Such replies land because familiarity and surprise work at the same time. Most users know their habits or weaknesses to some extent, but they do not often face them condensed into a one-line insult. ChatGPT does not uncover new hidden secrets, but links the dots of information the user has already provided and summarises them in an unexpected way.

When the user asked for the most severe put-down as a final request, ChatGPT replied, "Your to-do list is meaningless now." It added, "Future historians will discover piles of neatly organised lists for things you were definitely going to do next month."

What emerged in the process was a new way of using chatbots. Users are starting to use chatbots as a mirror reflecting themselves, beyond information searches or work assistance. It is not a structured format like personality tests or self-help books, but short, aggressive feedback based on chat history is being received more directly.

Still, such put-downs only hold within the range of information the chatbot knows. ChatGPT merely builds jokes based on what users provide and repeated behavioural patterns, and it does not uncover hidden facts. Even so, when it reveals links users did not fully recognise, it can lead beyond entertainment to motivation.