Prompting Fundamentals for Writers

If you have ever typed "write me a blog post about marketing" into ChatGPT, you already know what bad prompts produce: bland, generic, slightly cringe-worthy output that needs a rewrite anyway. The fix is not a longer prompt or a smarter model — it is a structured prompt. In this lesson you will learn the simple four-part recipe that turns AI from a clueless intern into a thoughtful collaborator.

This is the single most leveraged lesson in the course. Master prompting and every other lesson becomes 3x easier.

What You'll Learn

• The four-part RCTO prompt formula every AI writer should know
• The five rookie prompt mistakes — and the simple fixes
• How to use system prompts and "act as" framing to lock in tone
• The "back-and-forth" pattern that produces the best AI writing

Why Most Prompts Fail

When you type "write me a blog post about marketing," you have given the AI:

• No audience (students? CMOs? small business owners?)
• No angle (101 explainer? contrarian opinion? case study?)
• No length (300 words? 2,000?)
• No tone (formal? funny? academic?)
• No format (listicle? narrative? interview?)
• No example of what good looks like

The AI fills every blank with the safest, most average answer it can. Result: corporate sludge.

The RCTO Formula

The four parts of every great writing prompt:

• R — Role. Who should the AI be? "Act as a senior copywriter who specializes in technical brands."
• C — Context. What's the situation? "I am a CS student writing my first portfolio blog post. My audience is recruiters at tech startups."
• T — Task. What exactly do you want? "Write a 700-word blog post explaining how I built a Twitter bot that detects spam."
• O — Output format. What should the result look like? "Use H2 subheadings, short paragraphs, one code snippet, and a one-line takeaway at the end."

Compare the bad prompt to the RCTO version:

Bad: "write me a blog post about my Twitter bot project"

Good (RCTO):

Act as a senior copywriter who writes for technical audiences. I am a CS student writing my first portfolio blog post. My audience is recruiters at tech startups who skim before deciding whether to interview me. Write a 700-word blog post explaining how I built a Twitter bot that detects spam, including the technical decisions I made. Use H2 subheadings, short paragraphs (3 sentences max), one code snippet, and a one-line takeaway at the end. Tone: confident but not arrogant.

The second prompt produces something you would actually publish. The first produces sludge.

The Five Rookie Mistakes (and the Fix)

Mistake 1: No audience. Fix: always name the reader. "for a 19-year-old student" produces wildly different prose than "for a tenured economics professor."

Mistake 2: No length cap. Fix: specify words or sentences. "in 250 words," "in three short paragraphs," "in one tweet under 280 characters."

Mistake 3: No tone instruction. Fix: pick two or three adjectives. "warm and slightly nerdy," "professional but conversational," "blunt, no marketing fluff."

Mistake 4: No example. Fix: paste a sample of writing you love and say "match this voice." This is the single highest-impact tip in the lesson.

Mistake 5: One-shot expectation. Fix: do not expect the first answer to be perfect. Iterate. The next section explains how.

The "Back-and-Forth" Pattern

Beginners write one big prompt and accept the first output. Pros treat AI as a conversation. Try this five-message ritual:

1. Frame the goal. "I want to write a 600-word blog post about my summer internship at a startup. The goal is to attract recruiter attention."
2. Have AI ask you questions. "Before drafting, ask me five questions to get the right details out of me."
3. Answer the questions. Real, specific answers — not generic ones.
4. Get the draft. "Now write the draft based on my answers."
5. Edit together. "Make paragraph 3 more concrete. Tighten the intro. Replace cliches."

Five messages, twelve minutes, and you have a publishable post. Compare that to the alternative — staring at a blank doc all evening.

Act-As Framing

The "act as" trick is short, specific, and powerful. Examples:

• "Act as a college admissions essay coach who has read 10,000 essays."
• "Act as a senior product marketer at Notion."
• "Act as a no-nonsense newspaper editor with a 500-word limit."
• "Act as my brutally honest friend who is also a startup founder."

Each role pulls the model into a different prose style. It is the cheapest tone-control trick that exists.

Using Examples (Few-Shot)

Instead of describing the style you want, show it. Paste a paragraph you love and write:

Match this voice in everything I ask next. Voice sample: [paste paragraph here]

Now write a 200-word LinkedIn post announcing I just finished a Python course on FreeAcademy.ai.

The AI will study the rhythm, vocabulary, sentence length, and quirks of your sample and try to mirror them. This is the secret behind blogs that sound human.

Constraints Are Your Friend

Counterintuitively, more constraints produce better writing, not worse. Try:

• "No adjectives in the first paragraph."
• "Every sentence under 15 words."
• "Use the word 'simple' exactly once."
• "Open with a question."
• "End with a one-line punchline."

Constraints force creative choices the model would otherwise default away from. Pile on three or four per prompt.

The "What's Wrong With This?" Move

After AI gives you a draft, the highest-leverage follow-up is:

Critique this draft as if you were a senior editor at The New Yorker. List the three biggest weaknesses. Then rewrite the draft addressing them.

The AI will brutally roast its own first answer and produce a much better second draft. Use this every time.

A Quick Practice Exercise

Open Claude (claude.ai) and run this exact prompt:

Act as a writing coach. I want to write a 300-word LinkedIn post celebrating that I just finished a free AI course on FreeAcademy.ai. My audience is recruiters in tech. Tone: confident, specific, no humble-brag. Before drafting, ask me five short questions to get the details right. Wait for my answers.

Answer the questions for real. Ask Claude to write the draft. Then say:

Critique that draft like a senior LinkedIn editor. List three weaknesses, then rewrite.

Save the final result. You may use it later in the course.

Key Takeaways

• The RCTO formula — Role, Context, Task, Output — is the four-part recipe that turns AI from clueless to thoughtful.
• Most bad AI writing comes from prompts missing audience, length, tone, or examples. Add them.
• Treat AI as a conversation, not a one-shot vending machine. The five-message ritual produces dramatically better drafts.
• "Act as" framing and pasting voice samples are the two cheapest tone-control tricks that exist.
• After every draft, ask AI to critique itself and rewrite. The second draft is almost always better than the first.