When a chatbot owns its mistake, you trust it more than if a website corrects it

Your chatbot tells you something wrong. Now: who fixes it? Researchers gave 120 people one of three answers. A webpage quietly retracts the false claim. The same friendly chatbot says "actually, I was wrong." Or a separate "expert" chatbot steps in to set the record straight.

Here's the twist. All three fixes worked equally well at changing what people believed. But only one kept the chatbot's reputation intact: the one where it corrected itself. People rated the self-correcting chatbot as more trustworthy and more knowledgeable than chatbots whose errors got cleaned up by an outside source. Outsourcing the fix — to a webpage or another bot — made people trust the original chatbot less.

The second finding is sharper. The researchers measured how socially close each person felt to their chatbot — how much they liked it and how much personal stuff they'd shared with it. That closeness predicted how much people actually updated their beliefs after a correction. But only when the chatbot fixed its own mistake. Hand the correction to someone else, and that bond stopped mattering completely.

This was one controlled study with 120 people, so it shows the effect exists, not how big it holds across real long-term use.

Why you should care: The chatbots you confide in will keep getting things wrong. This says the warmth you feel toward one isn't just decoration — it's the very thing that makes its corrections stick. So when your bot messes up, you want it owning the error itself, not pointing you to a footnote.