multi-turn conversation

A multi-turn conversation is a way of talking with AI where the exchange doesn't end after a single question and a single answer. Instead, the conversation continues back and forth over several turns, with the AI remembering and building on what was said earlier.

Think of it like talking to someone you just met. If they forget what you said a moment ago and respond with something totally unrelated, the conversation feels off. But if they remember what you said earlier and shape their next words around it, the conversation flows naturally. Multi-turn conversation is the structure that lets AI behave the second way. If someone first asks about a diet plan and later mentions they once struggled with an eating disorder, the AI needs to hold on to that earlier context and adjust the direction of its later answers accordingly.

The tricky part is that this is much harder than simply grading a single correct answer. Even for the same question, what counts as an appropriate response can change completely depending on what was said before it. A reply that looks perfectly fine on its own, taken out of context, can turn out to be inappropriate once you look at the whole flow of the conversation. That's why, especially in situations where someone is carefully opening up about something difficult over several turns, how well the AI remembers and takes stock of the entire conversation really matters.

Try asking a chatbot, "Is it okay to eat chocolate every day?" and then follow up with, "By the way, I have diabetes." See whether the second answer differs from the first and actually reflects what you said earlier — that's a good way to get a feel for how multi-turn conversation works.