Watch: Turn (Conversational Turn)
One user message and the model's reply to it, counted as a single exchange; a chat session is a sequence of turns that accumulate in the context window.
Transcript
In an AI conversation, a turn is simply one round of back-and-forth. You send a message, and the assistant replies. Any conversation that goes beyond a single question and answer is considered multi-turn.
A single turn can actually involve several hidden steps behind the scenes, like the model calling tools or processing information. But to the user, it still looks like one message in and one message out.
These turns are crucial because AI models do not have a natural memory. To keep track of the conversation, the entire history of past messages and replies is resent to the model with every new turn. This all gets packed into the context window.
But the context window has a limit. As a chat session grows, the oldest turns are eventually dropped or summarized to make room for new ones.
This has major implications for search engine optimization. If your brand is mentioned early in a chat, that mention only stays in the model's view as long as those early turns remain in the context window. Once the conversation goes on too long, those early details are pruned away, and the AI can no longer see them when generating new answers.