An AI prompt builder generates structured prompts for ChatGPT or Claude, optimized for sales outreach. Pick your goal, target, channel. Get a 7-part prompt that gets 2-3x better replies than "write me a cold email."
1 What's your goal?
2 Who are you targeting?
3 What channel?
4 Your product / service
5 Industry (optional)
Fine-tune (optional)
This uses the 7-part expert framework: Task â Context â Reference (good + bad example) â Success Brief (how the reader should feel) â Rules â Conversation (AI asks before writing) â Output format. Each part makes the output dramatically better. Claude prompts use XML tags for maximum precision.
Define the task and target. Tell the AI what type of email (cold email, follow-up, LinkedIn DM) and who the recipient is (VP Sales, CEO, CTO). Generic prompts produce generic emails.
Add constraints and reference examples. Include a good example and a bad example. Set word count (50-125 for cold email). Ban spam words. These constraints separate a 2% reply rate from 12%.
Use the 7-part framework: Task, Context, Reference, Success Brief, Rules, Conversation, Output. This forces the AI to understand what success looks like. The Conversation step â where the AI proposes its angle before writing â is the quality lever most people skip.
Cold email benchmarks: what the data says
Structured vs generic prompts: Emails from the 7-part framework get 2-3x higher reply rates than "write me a cold email" prompts.
Optimal length: 50-125 words. Under 50 lacks context. Over 125, reply rates drop 40%.
Question CTAs vs statements: Ending with a question increases replies by 14%.
Subject lines: Under 5 words, lowercase, outperform longer title-case subjects.
Methodology: Based on testing 500+ prompt variations across 200+ B2B outbound campaigns run through Overloop.
Why structured prompts write better cold emails
1
7-part expert framework
Task, Context, Reference, Success Brief, Rules, Conversation, Output. Each section makes the AI output dramatically better. This is how prompt engineers actually work.
2
Good + bad examples baked in
Every prompt includes a reference email the AI should match AND a bad example to avoid. Few-shot examples are the single biggest quality lever in prompting.
3
AI asks before writing
The prompt tells the AI to propose its angle and wait for your approval. You stay in control. The AI collaborates instead of guessing.
Questions we get every day
Claude produces better cold email tone out of the box â casual and professional without heavy constraint-stacking. ChatGPT is better for research and data extraction. The best approach: use ChatGPT for prospect research, Claude for writing the actual email. This prompt works with both.
Because AI models optimize for whatever you ask. "Write a cold email" = generic, formal, long. A structured prompt with role, constraints, and specific instructions = sharp, short, human. The difference in output quality is 10x. The prompt IS the product.
10, not 35. One per channel (cold email, LinkedIn connect, LinkedIn DM, follow-up, breakup) Ã 2 goals. This builder generates the exact prompt for each combination. Quality over quantity.
It's the structure that gets the best output from AI. Role = who the AI is pretending to be. Context = background on the situation. Task = what to produce. Constraints = rules for the output. Without this structure, AI defaults to generic, formal, and long â the opposite of what works in cold email.
Yes. Build and copy as many prompts as you want. No signup, no limits.