Prompting for Rough Drafts

Version 2.0 Standard: Premium

AI can produce a passable rough draft in 2 minutes. Without the right prompting, that draft will take you 2 hours to fix. The quality of an AI draft is almost entirely determined by the quality of your prompt — and most prompts are too vague, too broad, or too trusting. This lesson teaches you how to prompt for drafts that require editing, not rewriting.

---

Part 1 — The Draft Spectrum: Bad to Usable

What AI Drafts Look Like at Each Prompt Quality Level

Prompt Quality	Draft Output	Your Editing Time
No prompt (just a topic)	Generic, thin, interchangeable with any competitor	3–4 hours (essentially a rewrite)
Basic prompt (topic + audience)	Better targeted, still surface-level	2–3 hours
Structured prompt (topic + audience + angle + constraints)	Solid structure, needs voice and depth	1–2 hours
Expert prompt (full briefing with outline, examples, and anti-patterns)	Near-publishable draft requiring polish	30–60 minutes

Your Goal

Move every draft to the "Expert prompt" tier. The 15 minutes you spend crafting the prompt saves 2+ hours of editing. This is the highest-ROI activity in AI-assisted writing.

---

Part 2 — The Section-by-Section Drafting Method

Why You Should Never Ask AI to Write the Whole Article at Once

The Problem

When you prompt: "Write a 2,000-word article about email marketing automation," AI:

• Loses quality after ~800 words — later sections get thinner and more repetitive
• Runs out of unique points — starts restating earlier ideas in different words
• Cannot maintain tone — the voice shifts subtly across long outputs
• Ignores your angle — starts specific, defaults to generic by the end

The Solution

Draft one section at a time. Each prompt focuses on one H2 section with full context.

The Section Prompt Template:

Section to draft: [H2 heading] Context: This is section 3 of 7 in an article about [topic]. Previous sections covered [brief summary]. Next section will cover [brief summary]. Audience: [who] Tone: [specific voice notes] Section requirements:

• Key points to cover: [list 3–5 bullet points]
• Include: [specific example, data point, or comparison]
• Avoid: [specific anti-patterns]
• Length: [target word count for this section, e.g., 200–300 words]

---

The Draft Assembly Process

flowchart TD
    A[Approved Outline\nH1 + H2s + H3s] --> B[Draft Section 1\nOne Prompt per Section]
    B --> C[Quick Review\n+ Edit Section 1]
    C --> D[Draft Section 2\nInclude Context from S1]
    D --> E[Quick Review\n+ Edit Section 2]
    E --> F[Continue Until\nAll Sections Drafted]
    F --> G[Full-Article Pass\nContinuity + Transitions]
    G --> H[Polish Pass\nVoice + Formatting]

    style A fill:#1A3557,color:#fff
    style H fill:#217346,color:#fff

---

Part 3 — Controlling Tone, Depth, and Style

The Three Knobs You Must Set

Tone Control

The problem: AI defaults to a formal, encyclopedic tone. Sentences like "It is widely recognized that..." and "One may consider..."

How to control it:

Instruction	AI Interprets As
"Write like a senior colleague explaining to a junior"	Conversational but professional. Avoids jargon without dumbing down
"Write like a blog post, not a textbook"	Shorter sentences. Contractions. More opinionated
"Match this writing sample: [paste 100 words]"	Mimics sentence length, vocabulary level, and personality
"No passive voice. No sentences over 20 words"	Direct, punchy, active

Depth Control

The problem: AI either skims the surface or pads with filler. Rarely hits the right depth.

How to control it:

Instruction	Effect
"Explain this to a beginner who has never encountered the concept"	Defines terms, uses analogies, avoids assumptions
"Explain this to a practitioner who wants actionable techniques"	Skips definitions, provides specific steps and tools
"For each point, include: the principle, a real-world example, and a common mistake"	Forces depth through structure
"After explaining each concept, provide a worked example"	Prevents surface-level coverage

Style Control

The problem: AI uses generic constructions: "It is important to note that..." / "There are several key factors..."

How to control it:

Anti-Pattern to Ban	Replacement Instruction
"It is important to note that..."	"State the point directly. No meta-commentary"
"There are several factors..."	"Name the exact number and list them"
"In today's digital landscape..."	"No throat-clearing. Start with the first useful sentence"
"In conclusion..."	"End with a specific action the reader should take"

---

Part 4 — Bad vs. Good Examples

❌ Vague Prompt → Bad Draft

Prompt: "Write a section about email subject lines."

AI output: "Email subject lines are the first thing recipients see when they receive your email. A good subject line can significantly improve your open rates. It is important to keep subject lines short, use personalization, and create a sense of urgency. Many experts recommend A/B testing your subject lines to see what works best for your audience."

(Why it fails: Generic, surface-level, no specifics. "Many experts recommend" — which experts? "Significantly improve" — by how much? The writer now has to research and rewrite from scratch.)

✅ Specific Prompt → Usable Draft

Prompt: "Draft a section titled 'The 3-Second Subject Line Test.' Audience: email marketers with 1+ years experience. Tone: senior colleague. Cover: why subject lines function as micro-hooks (link to the Hook Writing lesson), the 30-character mobile preview constraint, and one A/B test framework. Include specific numbers from Mailchimp's 2024 benchmark data. Anti-patterns: no 'it is important,' no passive voice, no sentences over 20 words. Length: 200–250 words."

AI output: "Your subject line has 3 seconds and 30 characters to earn an open. Not 60 characters — 30. That's what Gmail and Apple Mail show on mobile, where 68% of emails are read (Mailchimp, 2024).

Think of it as a micro-hook (if you completed the Hook Writing lesson, the same principles apply here). The subject line's job isn't to summarize the email — it's to create a reason to open it.

The fastest A/B test framework:

1. Write 5 subject line options
2. Filter to the 2 that pack the most curiosity into ≤30 characters
3. A/B test with 20% of your list for 2 hours
4. Send the winner to the remaining 80%

One pattern that consistently wins: specific numbers. 'Save 23% this week' beats 'Great savings inside' by an average of 14% in open rate (Mailchimp benchmark, 2024)."

(Why it wins: Specific, actionable, data-backed. Ready for light editing, not a rewrite.)

---

Part 5 — The Anti-Pattern Blocklist

Include this blocklist in every drafting prompt.

Add this to every prompt to prevent the most common AI failures:

**DO NOT USE:**
- "In today's digital landscape..."
- "It is important to note that..."
- "There are several key factors..."
- "In the ever-evolving world of..."
- "Studies show..." (name the specific study)
- "Experts agree..." (name the specific expert)
- Passive voice ("it was found that...")
- Sentences over 25 words
- Paragraphs over 4 sentences

---

Part 6 — Output Checklist

Before moving to the next lesson, confirm every item below.

• Section-by-section drafting: You never prompt for a full article — always one section at a time.
• Expert-level prompts: Your prompts include audience, tone, key points, examples, anti-patterns, and word count.
• Tone control: You set specific tone instructions, not just "professional" or "casual."
• Depth control: You specify the expertise level of the reader and require examples per concept.
• Anti-pattern blocklist: Every prompt includes the banned phrases list.
• Edit, don't rewrite: Your AI drafts require 30–60 minutes of editing, not 3 hours of rewriting.

---

Internal use only. Do not distribute externally. For questions or suggested updates, raise with the content lead.