Stockertim Dictionary · Artificial intelligence
Iterative prompting
Also called iterative refinement, prompt iteration
In short
Iterative prompting means improving an output over several rounds of targeted follow-up instructions, each aimed at one specific fault in the previous draft.
- Beginner
- Process
- Checked
In detail
The first answer from an AI assistant is a draft, not a finished piece. Iterative prompting improves it in rounds. After reading the draft, the user names one fault and asks for that to be fixed: the opening is too long, the second figure lacks a source, the tone is too formal for the audience. The next draft is read again and the cycle repeats until the output is good enough.
Targeting one or two faults per round works better than a vague instruction such as make it better, because the model knows exactly what to change and the user can see whether the change worked. A short change log, noting what was asked in each round, makes it easy to turn the final set of fixes into a better starting prompt next time.
Long rounds of revision can drift. Earlier constraints may be forgotten, or errors introduced in one round may persist. When that happens it is often quicker to start a fresh conversation with an improved prompt that includes the lessons learned.
Worked example
A marketing assistant at Northwind Bikes refines a product description over three rounds.
- Round 1 follow-up: "Cut the opening to one sentence; customers read this on a phone."
- Round 2 follow-up: "Replace the claim about weight with the figure in the spec sheet: 11.2 kg."
- Round 3 follow-up: "Change the tone from formal to friendly, keeping it under 80 words."
- Reason: each follow-up fixes one named fault, so every change is visible and checkable.
After three rounds the description is short, accurate and on tone, and the three fixes go into the team's starting prompt.
In practice
Staff use follow-ups to tighten drafts, fix facts and adjust tone, keep a note of the fixes, and fold them into templates so the next first draft starts closer.
Watch out
- Vague follow-ups such as make it better give unpredictable changes.
- Many rounds in one long chat can lose earlier instructions; restart when it drifts.
Learn it properly
Official sources
- Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile · National Institute of Standards and Technology
Facts last checked on 3 November 2026.
Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.