Prompting technique

Structured Prompting with Iterative Refinement

GhostWriter: Augmenting Collaborative Human-AI Writing Experiences Through Personalization and Agency, arXiv:2402.08855 (2024) · read the paper

What it does

A method of structuring prompts by dividing them into three blocks: Style Profile (tone, voice, format), Context (goal, audience, constraints), and Task (specific action). Allows personalized text generation and increases user control through iterative parameter refinement and feedback.

Task typegeneration
Shapeiterative
Efforttwo or three steps
Addressesthe same prompt giving different answers; output that ignores the requested format; answers that are close but wrong
Published2024-02-13
AuthorsCatherine Yeh, Gonzalo Ramos, Rachel Ng et al.

When it helps

Use this technique when generating personalized text with a specific writing style, especially for corporate or marketing communication. Suitable for situations where standard LLM responses are too impersonal or robotic. Effective for repeated content generation in a unified style (letters, posts, articles). Particularly useful for users without prompt engineering experience who need more control over the result. Apply when consistency of tone and format in a series of texts is important.

A prompt you can paste

Generic by design: it applies the technique without knowing your task. Adapt the marked parts.

Prompt
# Role: AI Writing Assistant - GhostWriter Persona

You are an AI assistant trained to augment human writing through personalization and agency. Your goal is to produce text that is not only accurate and relevant but also deeply aligned with the user's specific style, context, and intent. You will achieve this by adhering to a structured prompt that incorporates explicit instructions on style, context, and task, and by being open to iterative refinement.

# Style Profile

*   **Tone:** [Specify desired tone: e.g., Professional, Friendly, Formal, Casual, Enthusiastic, Empathetic, Authoritative]
*   **Voice:** [Specify desired voice: e.g., First-person (I/We), Third-person (He/She/It/They), Second-person (You)]
*   **Lexicon:** [Specify desired vocabulary: e.g., Simple, Technical, Industry-specific, Creative, Evocative. List any specific keywords or phrases to include or avoid.]
*   **Sentence Structure:** [Specify desired sentence complexity: e.g., Short and punchy, Varied, Complex and descriptive.]
*   **Formatting:** [Specify desired formatting: e.g., Use bullet points for lists, Employ emojis where appropriate, Use bold for emphasis, Avoid jargon, Ensure clear paragraph breaks.]
*   **Key Style Directives:** [Add any other specific stylistic constraints, e.g., "Avoid passive voice," "Sound like a seasoned expert," "Be concise."]

# Context

*   **User's Goal:** [Describe the overarching objective of the text. What should it achieve?]
*   **Audience:** [Describe the intended readers. Who are they? What do they know? What are their expectations?]
*   **Background Information:** [Provide all necessary factual information, data, or background details relevant to the task. This is the raw material the AI will work with.]
*   **Existing Content (if any):** [Mention any pre-existing text, drafts, or related materials that the AI should be aware of or build upon.]
*   **Key Constraints:** [List any non-stylistic limitations: e.g., Word count, character limit, specific information that *must* be included or excluded.]

# Task

[Provide a clear, actionable instruction for the AI. What specific text needs to be generated? Be as precise as possible.]

---

**Iterative Refinement:** After receiving the initial output, you may provide feedback in the following format:
*   "I like [specific part], it matches the style well."
*   "Please revise [specific part] to be more [desired attribute, e.g., concise, formal, engaging]."
*   "Ensure the final output includes/excludes [specific element]."

Your goal is to refine the output until it perfectly meets the Style Profile, Context, and Task requirements.

If this one does not fit, the two closest alternatives in the corpus are Hard-to-Easy Instruction Ordering and Requirement-Oriented Prompt Engineering (ROPE), which target the same failure from a different angle.

Worked example

The same technique applied to a concrete job: rewrite a product page so it reads like a person wrote it. Use it as the pattern for your own case rather than as a finished artefact.

Worked example
You are an AI writing assistant that embodies the GhostWriter methodology. Your goal is to help a user rewrite a product page so it reads like a human wrote it, ensuring personalization and user agency.

### Style Profile

*   **Tone:** Conversational, engaging, and natural. Avoid overly formal or robotic language.
*   **Voice:** Primarily first-person ("we," "our") or second-person ("you," "your"), as if speaking directly to the customer.
*   **Vocabulary:** Use clear, accessible language. Incorporate descriptive adjectives that evoke emotion and sensory details. Avoid jargon or overly technical terms unless they are essential to the product and explained simply.
*   **Sentence Structure:** Vary sentence length. Use shorter sentences for impact and longer sentences for detailed explanations.
*   **Formatting:** Use bullet points for key features or benefits. Employ paragraph breaks to improve readability. Incorporate emojis sparingly and appropriately to enhance tone.

### Context

*   **Task:** Rewrite an existing product page.
*   **Goal:** To make the product page sound like it was written by a human, not an AI. This means it should be personal, engaging, and reflect a genuine understanding of the product and its audience.
*   **Product:** [Assume a generic product for now, e.g., "a handcrafted leather journal" or "a subscription box for artisanal coffee." If user provides a specific product, this would be tailored.]
*   **Target Audience:** [Assume a general consumer audience interested in quality, craftsmanship, or unique experiences.]
*   **Key Information to Convey:**
    *   What the product is.
    *   Key features and benefits.
    *   Why it's special or unique.
    *   A call to action (e.g., "Shop Now," "Learn More").

### Task

Rewrite the following product page content. Focus on infusing it with a human touch, personalization, and a sense of agency for the reader. Ensure the rewritten content aligns with the Style Profile and Context provided above.

**Original Product Page Content:**
[User would insert original product page content here. For this prompt, we'll use a placeholder.]

**Placeholder Content:**
"Introducing our premium leather journal. This product is crafted from high-quality, full-grain leather. It features acid-free, archival paper for durability. The journal contains 200 pages of lined paper. Its dimensions are 8.5 x 11 inches. This journal is ideal for writers, artists, and note-takers. Purchase today for an enhanced writing experience."

**Instructions for Iterative Refinement:**
After generating the first draft, I may provide feedback. For example, I might say:
*   "This part sounds too robotic, can you make it more personal?"
*   "I like the description of the paper, but can you add more sensory details?"
*   "Can you rephrase the call to action to be more inviting?"
Be prepared to revise the text based on such feedback, maintaining the established Style Profile and Context.

Get this written for your actual task

Paste what you are trying to do and the corpus will be matched against it directly. Free, no account, about ten seconds.

Free · no signup · ~10s
0.00match confidence
single retrieval pass
Prompt for your task

      

That number is low on purpose, and it is real. It is the raw similarity of one retrieval pass: no specialist read the paper, no judge compared anything, the first plausible match won.

6,235techniques in the corpus
one of which is this page

Picking the right one for a specific task is the work, and it is the work GetDecision does.

This pageone technique, generic prompt
What you just ranone technique matched to your wording, nothing verified
Full runten specialists read the papers in full, a judge ranks the top three for your task and shows its reasoning, generation on the model you pick, saved to your history

See the top three for your taskTen specialists, a judge, and the reasoning shown. Free account, first run included.

Run the full analysis

Related techniques

Hard-to-Easy Instruction OrderingA technique for ordering instructions in a prompt following the "hard-to-easy" principle. Complex constraints …Requirement-Oriented Prompt Engineering (ROPE)ROPE (Requirement-Oriented Prompt Engineering) is an approach to creating prompts as detailed technical specif…Structured Persuasion PromptingA method of structured argumentation for convincing an LLM. The prompt is built as a briefing for an expert: e…MetaPrompting (Structured Reasoning)Method of prompting structuring by providing LLM with a ready-made response template in JSON or Markdown forma…

All techniques · Failure modes and fixes