What Is Prompt Chaining?

Prompt chaining is the technique of splitting one large, complex task into a sequence of smaller, focused prompts. The output of each step becomes the input to the next, forming a reliable pipeline instead of asking the model to do everything in a single overloaded prompt.

It is the foundation of most serious AI workflows. When a task has several distinct phases — research, structure, drafting, polishing — chaining keeps the model focused on one job at a time, which dramatically improves quality and control.

A Classic Example Chain

Here is a four-step chain that turns a raw document into a polished piece of writing.

The Chain in Practice

Each step is its own prompt. Notice how the output of one is pasted into the next.

Step 1 — Extract

Extract the 5 most important points from the text below.
Return them as a short bulleted list, nothing else.

"""
[paste source document here]
"""

Step 2 — Outline

Using the key points below, create a logical outline with
a clear introduction, 3 body sections, and a conclusion.

Key points:
[paste output from Step 1]

Step 3 — Draft

Expand the outline below into a full draft.
Write in a clear, professional tone, about 400 words.

Outline:
[paste output from Step 2]

Step 4 — Polish

Polish the draft below: improve flow, remove repetition,
fix grammar, and keep it under 400 words.
Return only the final version.

Draft:
[paste output from Step 3]

Automating a Chain (Optional)

In code, a chain is simply feeding each response into the next call. Here's the idea in pseudo-JavaScript:

// Prerequisite: an API client for your chosen model
async function runChain(sourceDoc) {
  const keyPoints = await ask(`Extract 5 key points:\n${sourceDoc}`);
  const outline   = await ask(`Create an outline:\n${keyPoints}`);
  const draft     = await ask(`Expand into a draft:\n${outline}`);
  const finalCopy = await ask(`Polish this draft:\n${draft}`);
  return finalCopy;
}

Single Prompt vs. Prompt Chain

Common Pitfalls to Watch Out For