Here’s a real AI cheat sheet. Six prompt patterns. They cover roughly 80% of what you’ll use AI for on any given workday. Not sixty. Not a hundred. Six.
I know that sounds too simple. But a Wharton study tested fancy prompting techniques against plain, clear instructions. On modern AI models, the fancy stuff barely moved the needle (2.9% improvement). On one model, it actually made things worse. The researchers’ conclusion: simple and clear beats clever and complex.
Meanwhile, 75% of knowledge workers already use AI at work. But only 39% have received any training. That’s the gap this cheat sheet fills. Not more techniques. Just the ones that actually matter.
These patterns work the same way in ChatGPT, Claude, Gemini, and pretty much any AI tool you’ll find on the AI market map. They’re not model-specific tricks. They’re how you talk to AI so it stops giving you generic answers.
Set the context (not just the role)
Every AI cheat sheet on the internet starts with “Act as a [role].” A marketing expert. A business consultant. A LinkedIn guru.
It feels like it should work. But Sander Schulhoff co-authored the largest academic study on prompting. 1,500+ papers, with researchers from OpenAI, Google, and Stanford. He tested 12 different role prompts on 2,000 questions. Role prompts barely changed accuracy. A prompt describing an “Idiot” actually outperformed a “Genius” prompt by 2.2 percentage points.
What does work is context. Tell the AI your situation, your audience, and your goal. Andrej Karpathy (who co-founded OpenAI and led AI at Tesla) calls this “context engineering.” It’s not about the perfect phrasing. It’s about what information you feed in.
Copy-paste this:
I'm a [your role] at a [company size/type].
I'm writing for [audience, be specific].
The goal is [what you want to achieve].
Here's what I know so far: [paste relevant background].Real example:
I'm the marketing lead at a 15-person B2B SaaS company.
I need a cold email for CTOs who tried [competitor] and dropped it.
The goal is to get a 15-minute call booked.
Keep it under 100 words. Our main selling point is faster onboarding.That produces something actually usable. Compare it to “You are an expert email copywriter. Write me a cold email.” Same tool. Wildly different output.
My take: I stopped using role prompts entirely about six months ago. I just describe my situation and what I need. The results got better, not worse. Roles change tone. Context changes substance.
If you’re picking which AI tools to use for marketing, this pattern works across all of them. Context is universal.
Show it examples
This is the single highest-impact technique in the academic research on prompting. Researchers call it “few-shot prompting,” which just means showing the AI a few examples before asking it to produce something.
The sweet spot is 2-3 examples. One example helps a lot. Three examples help more. Past five or six, the improvement flattens out.
Copy-paste this:
Here are two examples of the [style/format/tone] I want:
Example 1: [paste it]
Example 2: [paste it]
Now write one like these about [your topic].Real example for LinkedIn posts:
Here are two LinkedIn posts I like the style of:
Post 1: [paste a past post that performed well]
Post 2: [paste another one]
Write a new post about [topic] in this same voice.
Match the length, the paragraph breaks, and the level of detail.This works for emails, blog posts, product descriptions, ad copy, reports, and slide decks. Pretty much anything where you know what “good” looks like but don’t want to start from scratch.
When you’re building your full generative AI for marketing workflow, examples become your quality control. They replace long instructions about tone and style with a concrete target the AI can match.
Make it ask you questions first
This might be the most underrated pattern on any AI cheat sheet. Instead of trying to write the perfect prompt, flip it around. Let the AI figure out what it needs to know.
Ethan Mollick, a Wharton professor who studies AI at work, says to think of AI as “an infinitely patient coworker.” Coworkers ask questions before they start a task. Your AI should too.
Copy-paste this:
I need help with [task].
Before you start, ask me 5 questions that would
help you give me a much better answer.Real example:
I need to write a sales page for my new course on email marketing.
Before you start, ask me questions about the audience,
the course content, the price point, and what makes it different
from other email courses.The AI will ask things like: What level are your students? What’s the biggest objection? What results can they expect? These are questions you should have answered before writing, but probably hadn’t thought through yet.
The best part: it works for complex tasks where you don’t even know what a good prompt looks like. You don’t need to be a prompting expert. You just need to answer honestly.
My take: I use this pattern for anything where the stakes matter. Important emails, strategy docs, client proposals. The interview step takes 60 seconds and saves me from rewriting the output three times.
Ask it to poke holes
Most people use AI to make things. Write the email. Draft the plan. Create the outline. But AI is surprisingly good at a different job: finding what’s wrong with things you already made.
A BCG study with 758 consultants found something interesting. When professionals used AI for tasks it’s good at, they finished 12% more work, 25% faster, at 40% higher quality. But when they used AI for tasks outside its strengths? They performed worse than people without AI at all.
The lesson: don’t just hand AI a blank page. Hand it your work and ask it to find the weak spots.
Copy-paste this:
Here's my [plan/draft/strategy]:
[paste it]
What's wrong with this? What am I missing?
What would a skeptic say?Three variations that dig deeper:
Play devil's advocate on this marketing plan.
What are the 3 weakest assumptions?I'm about to send this email to 5,000 people.
What could go wrong? What would you change?Read this blog post draft. Where does it lose the reader?
Where am I making claims without evidence?This is where AI earns its keep for AI for entrepreneurs and marketers alike. You bring the domain knowledge. AI brings a second pair of eyes that never gets tired and doesn’t worry about hurting your feelings. If you’re evaluating a new venture, this critique pattern is exactly how to use AI for business ideas — stress-test the concept before you commit.
Tell it the format you want
Most generic AI output comes from not telling it the shape you want. “Write me a marketing plan” gets you 2,000 words of padding. “Give me a marketing plan as a table with 5 rows: channel, budget, expected result, timeline, first action” gets you something you can actually use in a meeting.
Copy-paste this:
Give me this as [format].
Keep it under [length].
Use [structure: headers, bullets, numbered list].Real examples:
Summarize this article as 5 bullet points,
each under 15 words.Compare these three tools in a table.
Columns: name, price, best for, biggest limitation.Write a 60-second elevator pitch.
Three paragraphs. Plain language. No jargon.You can stack constraints. “Under 200 words, in bullet points, written for someone who’s never heard of SEO” is a perfectly good instruction. The more specific your format, the less editing you do afterward.
This pattern pairs well with AI-enhanced content marketing. Once you know the format you want for each content type (blog intro, social post, email subject line), you can reuse the same format instructions across dozens of pieces.
If you’re building a full gen AI tech stack, format instructions become even more important. Different tools need different output shapes. Your email tool wants a subject line and body. Your CMS wants H2s and paragraphs. Your spreadsheet wants rows and columns. Same AI, different format instructions.
Chain your prompts (one job at a time)
Stop trying to do everything in one mega-prompt. A single prompt asking for “a complete marketing strategy with target audience analysis, competitive research, channel recommendations, and a 90-day plan” will give you something that looks complete. It’s actually shallow everywhere.
Instead, break it into steps. Each one builds on the last.
The pattern:
Step 1: "Research [topic]. Give me 10 key findings."
Step 2: "Based on those findings, outline a plan with 5 sections."
Step 3: "Write section 1 in detail. Keep it under 300 words."
Step 4: "Now review what you wrote. What's weak? Fix it."Real example for writing a blog post:
Prompt 1: "Here's my topic: [topic]. Give me 10 angles I could take."
Prompt 2: "I like angle 3. Build an outline with 6 sections."
Prompt 3: "Write the introduction. Start with a specific fact or story."
Prompt 4: "Read back what you've written. Is it too generic? Tighten it."This works because AI handles small, focused tasks much better than sprawling ones. Andrew Ng’s research showed that GPT-3.5 scored 48% accuracy on a coding benchmark with a single prompt. The same model hit 95% when the task was broken into steps.
My take: I chain prompts for anything that takes more than a paragraph to explain. Writing, research, strategy, even email sequences. The first prompt does the thinking. The second prompt does the building. The third prompt does the editing. Three conversations that each take two minutes beat one conversation that takes twenty.
If you want to explore the best AI platforms for business that support this kind of multi-step workflow, look for tools with conversation memory (ChatGPT, Claude) rather than single-shot generators.
How to use this AI cheat sheet in practice
You don’t need to use all six patterns on every task. The right pattern depends on the job.
For quick tasks (rewriting an email, summarizing a doc): context + format. That’s it. Takes 30 seconds.
For medium tasks (drafting a blog post, building a presentation): context + examples + format. Maybe three minutes of setup.
For high-stakes work (a strategy doc, a client proposal, a campaign plan): all six. Context, examples, interview, chain the steps, format the output, then ask it to poke holes at the end.
| Task | Patterns to use | Time to set up |
|---|---|---|
| Rewrite an email | Context + Format | 30 seconds |
| Write a LinkedIn post | Context + Examples + Format | 2 minutes |
| Draft a blog post | Context + Examples + Interview + Chain | 5 minutes |
| Build a marketing plan | All six | 10 minutes |
| Review a strategy doc | Context + Poke holes | 2 minutes |
McKinsey found that 78% of organizations use AI, but only about 6% are seeing real results. The gap is rarely the tool. It’s the skill. Salesforce data backs this up: 75% of marketers have adopted AI, but 84% still run generic campaigns.
That’s what this cheat sheet is really about. Not learning a hundred prompts. Learning the six moves that turn generic output into something you can actually use.
If you want a broader checklist for your whole team, the AI checklist for marketing teams covers all 15 steps (people, processes, and tools, not just prompts). For a full adoption roadmap, the AI adoption framework walks through the whole process.
And if you’re just getting started with AI in digital marketing, these six patterns are genuinely the place to begin. Mollick’s research shows that 10 hours of hands-on practice beats any formal training. Start with one pattern, use it for a week, then add the next.
For the full picture of implementing AI across your business, not just marketing, the same principles apply. Context in, examples in, break the work into steps. The patterns scale.
Build your own cheat sheet
The six patterns above are universal. They work for everyone. But the real leverage comes when you customize them for your specific job, your specific tools, and your specific workflow.
That’s what I do with founders and growth teams. We take these patterns and turn them into a system that fits how you actually work, not a generic PDF. If that sounds useful, you can see how it works on my work with me page.
FAQ
What is an AI cheat sheet?
An AI cheat sheet is a quick-reference guide to the prompt patterns that get the best results from AI tools like ChatGPT, Claude, and Gemini. Most AI cheat sheets you’ll find are 50-page PDFs stuffed with every technique imaginable. The useful ones are short (six patterns, not sixty) and focus on the techniques backed by research, not viral Twitter threads. Think of it as the moves you’ll actually reach for on a Tuesday morning, not homework you download and forget.
What’s the best ChatGPT prompts cheat sheet?
The one you’ll actually use. Six patterns (context, examples, interview, critique, format, chaining) cover about 80% of daily work. They work in ChatGPT and every other major AI tool. If you want prompts specific to marketing with AI, start with context and examples, because those two alone will change your output quality overnight. For marketing-specific AI prompts built on this structure, that guide covers the full breakdown.
What prompts should I know for AI?
Start with two: give it context (your situation, your audience, your goal) and tell it the format (table, bullet points, 100 words). Those two patterns alone fix most generic AI output. Once those feel natural, add “make it ask you questions first” for anything important. Research from Wharton shows simple, clear instructions beat elaborate frameworks.
Do I need to learn prompt engineering?
Not formally. The term “prompt engineering” makes it sound like a technical skill that requires courses and certifications. Mollick’s research shows that 10 hours of hands-on practice beats studying frameworks. Learn the six patterns in this cheat sheet and use them for a week. You’ll be ahead of most AI users who never received any training at all. If you’re curious whether AI marketing is actually worth it, the data says yes. But only if you learn to use it well.
Does this work for Claude, Gemini, and other AI tools, not just ChatGPT?
Yes. These are universal prompt patterns. They work because of how language models process information, not because of any ChatGPT-specific feature. I use them daily across ChatGPT, Claude, and Gemini. The patterns are the same. The only thing that changes is which tool handles your specific task best. You can explore that in the best AI for business guide.