AI Dictionary › Prompting
Multi-turn prompting is the technique of building a conversation with a model through several consecutive exchanges instead of entrusting the whole task to a single message. Each turn builds on the previous ones, allowing the user to correct, clarify or deepen the request as the model's response becomes visible.
Unlike a single monolithic prompt, here the user observes the intermediate output and decides how to proceed: asking for clarification, supplying a missing detail, correcting a misunderstanding or changing direction. The model keeps track of the exchange within its context window, so each new turn is interpreted in light of what was said before.
In modern AI, multi-turn prompting is the natural way conversational chatbots are used: starting with a broad request, refining the result with follow-up questions, asking to rewrite a section or change tone. It is especially useful for complex tasks such as drafting a long document, debugging code or exploring an idea, where the ideal answer emerges gradually through dialogue rather than in a single pass.
The term emerged alongside the spread of chat-based conversational models, when it became clear that multi-turn interaction produced more accurate results than a single prompt, precisely because it allows progressive refinement similar to a conversation between people.
Break the task into a sequence of exchanges instead of one giant prompt. Start by setting role and context, get a first draft, then refine with targeted follow-ups: "shorten section 2", "make the tone more formal", "add a counter-argument". Each turn inherits the full conversation, so corrections compound instead of starting over.
Turn 1: "You are an HR manager. Draft a job post for a junior data analyst, 150 words." Turn 2: "Good. Now remove the buzzwords and add the salary range 28-32k." Turn 3: "Rewrite the opening line so it speaks to career switchers."
Use it for any output you care about: drafts improve more per token than one-shot mega-prompts. Watch out for context drift on long chats: when the model starts contradicting earlier turns, summarize the state so far and continue in a fresh conversation.
JSON mode · Prompt scaffolding · Contrastive prompting · Context stuffing · Retry with backoff · Self-ask · Chain-of-thought · Role prompting · Prompt injection · Temperature · System prompt
Grace looks at whether, in multi-step scenarios, you use the follow-up turns to correct and refine the model's answer instead of starting from scratch each time.
From our network
Magellano GPS: Fleet Tracking Made Simple
Real-time GPS tracking, remote engine lock, fuel and CO₂ reporting for your fleet.
Visit magellanogps.com →From the Agora Intelligence blog
📱 Download the Android app (beta) iOS coming soon
Say what you mean. Get what you need.
Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.