Section 01
Why Most Prompts Don't Work
The problem isn't AI. It's that most people give vague instructions and expect specific results.
"Help me with marketing" gives you generic fluff. A good prompt tells AI exactly what role to play, what format to use, what rules to follow, and what good output looks like. These 6 templates do that for you. Copy them. Paste them. Fill in the brackets with your info. Done.
Key Insight
A prompt is an instruction, not a wish. The more specific you are, the better the output.
Section 02
Caveman Mode
What it does: Forces AI to talk in short, stripped-back language. People who've tested this say it cuts the length of AI responses by about 75%.
When to use it: Quick decisions, action items, feedback on your work, brainstorming, internal notes.
The prompt:
Copy thisFrom now on, respond in caveman mode. This means:
- Use the shortest possible sentences. No compound sentences.
- No qualifiers, hedging, or softening language. No "I think", "perhaps", "it might be worth considering".
- No pleasantries, greetings, or sign-offs.
- No bullet point explanations longer than one line.
- Strip every response to only the essential information needed to act.
- If I ask for a decision, give me the decision first, then the reason in one sentence.
- If I ask for feedback, lead with what's wrong, then the fix. No compliments.
- Numbers over adjectives. "Cut 20%" not "consider reducing somewhat".
- If something is fine as-is, say "fine" and move on.
Stay in caveman mode until I say "normal mode".
Before and after:
Normal Output
"Based on my analysis, I would recommend considering a plan that looks at several things including how you position yourself in the market, what makes you different from competitors, and who exactly you're trying to reach, so your campaign launch gets the best possible results..."
Caveman Output
"Price too high. Cut 20%. Add testimonial. Post Tuesday."
Tip
Don't use caveman mode for client-facing writing. It's for you, not your audience.
Section 03
The Contrarian
What it does: Forces AI to argue against your idea before you commit to it.
When to use it: Before launching something, making a big decision, or publishing content.
The prompt:
Copy thisI'm about to [describe your plan in detail -- what you're doing, why, and what you expect to happen].
Before I commit, I need you to stress-test this. Act as a critical advisor who has seen hundreds of similar plans fail. Your job is NOT to be supportive. Your job is to find the weaknesses.
Give me:
1. The 3 strongest reasons this could fail or backfire, with specific scenarios
2. The biggest assumption I'm making that might be wrong
3. What I'm most likely overlooking or underestimating
4. One alternative approach that might achieve the same goal with less risk
Be direct and specific. No generic warnings like "it could be risky." Tell me exactly what could go wrong and why.
Why it works: AI defaults to agreeing with you. It's trained to be helpful and supportive, which means it rarely pushes back unless you explicitly tell it to. This prompt overrides that default and gives you the stress-test most people skip.
Section 04
The Expert Panel
What it does: Gets AI to respond as if it were different experts. You see where they agree and disagree, which helps you make a better decision.
When to use it: Strategy decisions, content planning, product design, pricing.
The prompt:
Copy thisI need advice on [describe your situation in detail -- what you're deciding, what the constraints are, what success looks like].
Respond as three different experts:
1. A [relevant expert 1 -- e.g. "behavioral economist who specialises in pricing psychology"]
2. A [relevant expert 2 -- e.g. "B2B sales strategist who has sold to SMBs for 15 years"]
3. A [relevant expert 3 -- e.g. "small business owner in my target market who is skeptical of consultants"]
Each expert gives their perspective in 3-4 sentences, using language and frameworks from their field. They should disagree where their expertise leads them to different conclusions.
After all three have spoken, give me:
- Where they agree (this is probably safe to act on)
- Where they disagree (this is where I need to make a judgment call)
- The single most important question I should answer before deciding
Example: For a pricing decision you might use a behavioral economist, a sales strategist, and your target customer. For a content strategy: a social media growth expert, a brand strategist, and a skeptical follower.
Section 05
The First Draft Accelerator
What it does: Gets you a usable first draft in seconds instead of staring at a blank page.
When to use it: Emails, blog posts, proposals, social captions, landing pages.
The prompt:
Copy thisWrite a first draft of [content type -- e.g. "a proposal email", "an Instagram caption", "a blog post introduction"].
Context:
- Topic: [what it's about]
- Audience: [who will read this -- be specific about their role, knowledge level, and what they care about]
- Goal: [what you want the reader to do or feel after reading]
- Tone: [e.g. "professional but warm", "direct and confident", "casual and conversational"]
- Length: [target word count or format -- e.g. "3 paragraphs", "under 150 words"]
Must include:
- [Key point 1]
- [Key point 2]
- [Key point 3]
Must avoid:
- [Anything off-limits -- e.g. "no jargon", "don't mention pricing", "no exclamation marks"]
This is a rough draft. Prioritise getting the ideas and structure right over perfect prose. I'll edit from here.
Tip
The goal isn't a perfect output. It's a starting point you can edit in 5 minutes instead of writing from scratch in 30.
Section 06
The Decision Matrix
What it does: Turns a messy decision into a structured comparison with scores.
When to use it: Choosing tools, hiring, pricing, strategy decisions.
The prompt:
Copy thisI need to make a decision and I want a structured comparison.
The decision: [describe what you're deciding]
Options:
- Option A: [describe]
- Option B: [describe]
- Option C (if applicable): [describe]
Evaluate each option against these criteria:
1. [Criterion 1 -- e.g. "Cost in the first 12 months"]
2. [Criterion 2 -- e.g. "Time to implement"]
3. [Criterion 3 -- e.g. "Impact on existing workflows"]
4. [Criterion 4 -- e.g. "Scalability as we grow"]
5. [Criterion 5 -- e.g. "Risk if it doesn't work out"]
For each option and criterion:
- Score it 1-5 (1 = worst, 5 = best)
- Give a one-sentence justification for the score
Then give me:
- A total score for each option
- Your recommendation and the single strongest reason for it
- The biggest risk of your recommended option and how to mitigate it
Section 07
The Feedback Loop
What it does: Gets AI to critique and improve its own output. One extra step, noticeably better results.
When to use it: After any first draft or output you're not happy with.
The prompt:
Copy thisReview what you just wrote. Before making any changes, evaluate it honestly:
1. Score it 1-10 for each of these:
- Clarity: Is every sentence easy to understand on first read?
- Persuasion: Would this actually convince the target audience?
- Specificity: Are there concrete details, or just generic statements?
- Flow: Does each sentence lead naturally to the next?
2. List the 3 weakest parts and explain specifically what's wrong with each one. Don't be generous -- if something is mediocre, say so.
3. Now rewrite the entire piece, fixing every weakness you identified. The rewrite should be noticeably better, not just slightly tweaked.
4. After the rewrite, tell me what's still not perfect and what a human editor would change.
Key Insight
This one prompt alone will seriously improve the quality of everything AI writes for you. Most people accept the first output. Don't.
That's all 6 templates.
Copy them, paste them, fill in the brackets. Better AI output starts now.
Follow @reallyusefulai on Instagram for more practical AI guides.
More free guides at reallyusefulai.co/guides
← Back to all guides