You know the feeling. You open ChatGPT or Claude or Gemini, type a question, and get back something so generic it could have been written by a bored intern who read one Wikipedia article. So you try again with more words, and now it’s a wall of corporate-speak. You ask it to be simpler, and it turns into a greeting card. After three rounds you give up and write the thing yourself.
I spent a year and a half working on prompt engineering, and here’s what I eventually learned: the problem was never the AI. The problem was that I didn’t know how to talk to it. Prompting isn’t magic. It’s a skill with a structure, and once you learn the structure, the same template that gets you a great blog outline also gets you a great business plan, a great email, or a great piece of code. This playbook is the exact set of templates I use now, distilled into twelve prompts you can copy, paste, and adapt. By the end, you’ll understand why they work — and you’ll never accept a generic AI answer again.
WHY MOST PROMPTS FAIL (AND IT’S NOT WHAT YOU THINK)
Let me clear up the most common misconception first. People assume bad AI output comes from the AI being dumb. It’s not. These models are statistical prediction engines. Given a sentence, they predict the most likely next word. When you give them a vague prompt, they produce the most statistically likely vague response. That’s not a bug. That’s the model doing exactly what it was told.
The second reason prompts fail is context starvation. AI models are like a brilliant consultant who was hired but never briefed. If you walk into a meeting and say “write me a marketing plan,” the consultant will produce a generic plan because they have no information about your product, your audience, or your budget. But give them the full brief — who you sell to, what makes you different, what you tried before — and suddenly the output is worth something. Most people never give the brief.
The third failure is asking for the wrong thing. We ask for “a blog post” when we actually want “an outline we can react to.” We ask for “ideas” when we want “a ranked list with pros and cons.” The AI is happy to comply, but what comes back doesn’t serve us. Templates fix this because they force you to be specific about what you actually want.
THE FOUR-PART PROMPT STRUCTURE
Every effective prompt I’ve ever written follows the same four-part structure. I call it the SCOT framework: Situation, Context, Output, Tone. Once you can build a prompt from these four blocks, you don’t need templates anymore. But I’ll give you both.
Situation is who you are and what you’re doing. “I’m a solo freelancer writing content for small law firms.” This tells the AI your perspective and constraints.
Context is the material the AI needs to work with. This is where you paste your notes, your target audience description, your competitor list, your raw data. Context is the single most powerful ingredient, and it’s the one people skip most often.
Output is the exact form you want the answer in. “Give me ten headline options, each under 55 characters, formatted as a bulleted list, ranked from most clickable to least.” The more precise the output spec, the better the result. This is where templates do the heavy lifting.
Tone is how it should sound. “Write like a friendly expert, not a salesman. Short sentences. No corporate jargon.” Give the AI two examples of the voice you want if you can. Models copy style from examples far better than from adjectives.
Now let me put this structure to work. Here are the twelve templates.
THE TWELVE TEMPLATES
TEMPLATE 1: THE PERFECT BLOG POST OUTLINE
Use this when you want a structure you can actually write from, not a generic “intro, body, conclusion” skeleton.
Prompt: “Situation: I’m a blogger writing for [audience]. Context: My post is about [topic]. Here’s what I want to cover: [3-5 bullet points]. Here’s an example of an article I like: [paste URL or snippet]. Output: Create a detailed blog outline with an H1, a hook intro I can react to, H2 and H3 headings, and under each heading a one-sentence summary of the point plus a note on what data or example to include. Make the intro answer the reader’s core question in the first 50 words. Tone: direct, practical, no fluff.”
TEMPLATE 2: THE FEATURE-COMPARISON TABLE
Use this when you’re comparing tools, products, or approaches. It’s the most underrated template in existence.
Prompt: “Situation: I’m researching [product A vs product B] for [audience]. Context: I need to compare [features/pricing/support]. Here’s what I know: [your notes]. Output: Create a comparison table with columns for [feature, product A, product B, best for]. Below the table, write a paragraph explaining which option wins in each scenario. Tone: neutral, factual, helpful.”
TEMPLATE 3: THE COLD EMAIL THAT GETS REPLIES
The trick here is giving the AI the recipient’s world, not just your pitch.
Prompt: “Situation: I’m [name/role] reaching out to [recipient role] at [company]. Context: They [what they care about / what I know about them]. I can help them by [your value]. My relevant experience: [1-2 lines]. Output: Write a cold email under 120 words. Subject line under 6 words. Body: one line on why I’m writing, one line of proof, one clear ask. No flattery, no ‘I hope this email finds you well.’ Tone: confident, concise, human.”
TEMPLATE 4: THE CONTENT REPURPOSING MACHINE
Turn one piece of content into five formats. This template is pure leverage.
Prompt: “Situation: I created this content and want to maximize its value: [paste content]. Context: My audience is [audience] and they consume [channels]. Output: Repurpose this into 1) a 30-second LinkedIn post, 2) a 3-tweet thread, 3) a 5-point email newsletter, 4) a 60-second video script, 5) an FAQ of the three most likely follow-up questions. Tone: match the original, keep it punchy.”
TEMPLATE 5: THE CRITIC / RED TEAM
Use this to find the holes in your own thinking, business plan, or article before you publish.
Prompt: “Situation: I’m about to [launch this plan / publish this article / make this decision]. Context: Here is my plan: [paste]. Output: Act as a tough but fair critic. List the five biggest weaknesses, the top three risks, the one assumption most likely to be wrong, and one question I haven’t asked myself. Then give me a revised version that fixes the top two weaknesses. Tone: direct, constructive, no sugarcoating.”
TEMPLATE 6: THE STEP-BY-STEP LEARNER
Use this when you want to actually learn something, not just get an answer.
Prompt: “Situation: I want to learn [topic/skill]. My current level is [beginner / some experience]. Context: I have [time per week / tools available]. I learn best by [doing / reading / examples]. Output: Create a 30-day learning plan with a weekly goal, a daily task under 30 minutes, and a small project at the end of each week to test what I learned. Explain concepts in plain language and give one concrete example for each concept. Tone: like a patient mentor.”
TEMPLATE 7: THE RESEARCH SUMMARIZER
Use this when you’ve collected a pile of notes, articles, or data and need the signal.
Prompt: “Situation: I’m researching [topic] and gathered these sources: [paste notes/quotes/links]. Context: I need to decide [the decision the research informs]. Output: Summarize the key findings in 5 bullets, highlight where the sources disagree, flag anything that looks like bias, and give me the single most important takeaway for my decision. Tone: analytical, honest about uncertainty.”
TEMPLATE 8: THE EMAIL NEGOTIATION DRAFT
Use this for anything sensitive where words matter — salary, pricing, deadlines.
Prompt: “Situation: I’m negotiating with [person] about [issue]. Context: My position: [your ideal outcome, your walk-away point]. Their likely position: [what you know]. Our relationship: [working relationship]. Output: Draft a negotiation email in 3 versions: cooperative, firm but fair, and ready-to-walk-away. Each under 100 words. Give me the one sentence I should not say. Tone: professional, calm, respectful.”
TEMPLATE 9: THE “EXPLAIN TO A FRIEND” PARAPHRASER
Use this when an expert explanation leaves you more confused than before.
Prompt: “Situation: I’m trying to understand [concept]. Context: Here is the original explanation: [paste]. I got stuck at [the confusing part]. Output: Explain it again in three different ways: one with an analogy from everyday life, one with a simple concrete example, and one as a three-sentence version a 12-year-old could understand. Tone: warm, patient, zero condescension.”
TEMPLATE 10: THE CREATIVE BRAINSTORM WITH CONSTRAINTS
The secret to good AI brainstorming is constraints. Without them you get forty ideas, all bad.
Prompt: “Situation: I need [N] ideas for [goal — e.g., blog post topics, product names, marketing angles]. Context: My audience is [audience]. What I’ve already tried: [list]. What I won’t do: [constraints]. Output: Give me [N] ideas ranked by [impact / originality / feasibility]. For each: one-line description, why it works, and the biggest risk. Push me past obvious ideas — I’d rather have 3 weird ones than 10 safe ones. Tone: imaginative but practical.”
TEMPLATE 11: THE PRODUCT BRIEF WRITER
Use this when you need specs, feature descriptions, or a product page that converts.
Prompt: “Situation: I’m launching [product/service]. Context: What it does: [details]. Who it’s for: [audience]. Why it’s different: [differentiator]. Price: [price]. Output: Write 1) a product description under 80 words, 2) a feature list in order of benefit, 3) an answer to the top objection, and 4) a short version for a landing page hero. Tone: benefit-focused, specific, no hype.”
TEMPLATE 12: THE WEEKLY PLANNER
Use this on Sunday night to make Monday morning painless.
Prompt: “Situation: I’m planning my week. Context: My priorities are [top 3 goals]. My commitments: [meetings/events]. What I want to spend time on: [areas]. My tendency: [e.g., to overcommit, to skip deep work]. Output: Create a daily plan Monday-Friday with 1 deep-work block, a list of must-dos for the week, and a rule I should follow to protect my focus. Flag where my plan is unrealistic. Tone: supportive, direct.”
THE MISTAKES THAT SILENTLY KILL YOUR OUTPUT
Templates fix structure, but there are still a few habits that quietly destroy quality. I want to name them so you can avoid them.
Multitasking your request. Asking for an outline, a title, a meta description, and an email version in one prompt sounds efficient, but the AI splits its attention and every piece comes out weaker. Run one focused request per prompt.
No examples. Telling the AI “make it good” is useless. Showing it an example of good is worth a thousand adjectives. Find one piece of writing in the style you want and paste it in. The output quality jump is almost instant.
Accepting the first draft. The first draft is the AI’s guess at what you want. Feedback loops are where the magic happens. “Make it more specific. Add data. Cut the adjectives. Give me a version for a skeptical reader.” Three rounds of feedback beat ten carefully-worded first prompts.
Forgetting the human review. Prompt engineering is not a substitute for judgment. These tools are brilliant at drafting, summarizing, and structuring. They are mediocre at deciding what’s true, what’s tasteful, and what fits your voice. You’re the editor. The AI is the brilliant but undirected intern.
HOW TO BUILD YOUR OWN TEMPLATES
Once you’ve used these twelve, you’ll start seeing the pattern everywhere. Your own prompts will follow the SCOT structure automatically. Here’s my advice for building a personal library.
Save the templates that work. When you write a prompt that produces something you’d happily publish, copy the prompt into a notes file with a comment about what you used it for. In six months you’ll have a personal playbook that’s better than anything a course could sell you.
Build a context bank. Keep a file with your standard context blocks: your audience description, your bio, your product details, your writing style guide. Then every prompt you write starts by pasting in the relevant block. This one habit — pre-briefing the AI — will do more for your output quality than any other tip in this article. For more ready-made starting points, browse our collection of 50 ready-to-use prompts or the ten ChatGPT prompts we recommend for content creators.
Iterate in public. Share your best prompts with your team or your readers. Teaching prompt engineering is one of the best ways to learn it, and sharing prompts builds community. The people who get the most out of AI are the ones who treat prompting as a craft, not a party trick.
PUT IT TOGETHER
Here’s a quick way to start right now. Open your AI tool of choice. Take the thing you’ve been putting off — that blog post, that email, that research pile — and build a prompt using the SCOT structure. Situation, Context, Output, Tone. Be shameless about the detail. Paste in your notes. Specify the format exactly. Then ask for a first draft, give it feedback twice, and see what comes back. If ChatGPT is your tool of choice, our guide to its advanced techniques is the perfect companion.
I’ll bet my year of struggling that you’ll be shocked by the difference. The AI didn’t get smarter. You finally learned how to talk to it. That’s what prompt engineering really is: the skill of giving a brilliant, briefless consultant everything they need to do their best work. Learn it once, and every AI tool you ever touch gets better.
