Most cold emails get deleted in seconds because they smell fake, but this specific approach is reportedly achieving a massive 15-20% reply rate.
I recently found a fascinating breakdown by a Reddit user who spent months testing AI-generated cold emails. This expert didn’t just ask ChatGPT to "write an email"; the author ran a controlled test across more than 200 distinct outreach attempts to see what actually works. The findings were surprising. The creator realized that the biggest problem with AI writing isn’t that it’s too stupid, but that it tries too hard to be polite. It ends up sounding like a desperate salesperson rather than a helpful human. I was impressed by how simple the fix was.
💡 The "3rd Email of the Day" Rule
The core innovation this industry pro shared is a specific instruction: tell the AI to write as if it is the "3rd or 4th email someone has sent that day."
Think about how you write to a colleague or a client you talk to all the time. You don’t start with "I hope this email finds you well and you are having a magnificent Tuesday." You don’t use five adjectives to describe a simple report. You just say, "Hey, here is that file you needed, let me know if the data looks right."
This instruction strips away the fluff. It forces the AI to be concise, direct, and assume a level of peer-to-peer respect rather than looking for approval. When you write like you are already in a conversation, you bypass the recipient’s natural defense mechanisms against cold sales pitches. The author found that this single instruction kills the desperate tone that most Large Language Models default to.
📌 Why This Approach Wins
The Power of Negative Constraints
The original poster included a strict list of what the AI is not allowed to do. This is often more powerful than telling it what to do. The author explicitly bans phrases like "just reaching out," "I’d love to," or "wondering if." I think this is genius because these are "uncertainty markers." They signal to the reader that you are asking for a favor or that you are unsure of your own value. By removing them, the tone automatically shifts to one of confidence. The expert also banned exclamation points entirely. This keeps the tone calm and professional, preventing that fake, hyper-energetic "marketing voice" that turns people off immediately.
Context is the Price of Entry
One of the smartest parts of this setup is that the prompt is designed to fail if you don’t do your homework. The creator built the prompt to require specific inputs: the recipient’s role, recent news about their company, and a clear value proposition. It effectively refuses to write the email until it has these details. This prevents the user from lazy spamming. The prompt forces you to find a "trigger event": something specific and recent about the company, to use as the opening line. This proves to the reader instantly that this isn’t a blast sent to 10,000 people. It anchors the message in reality before you ever make an ask. It makes the email feel bespoke, even if the structure is automated.
The Low-Friction Ask
The final piece of the puzzle that the Reddit user highlighted is the "low-commitment ask." The prompt instructs the AI to end with a specific question or a single resource to review. It explicitly avoids big requests like "let’s jump on a 30-minute demo" or "can I pick your brain." By lowering the barrier to entry, the recipient feels safer replying. It turns the interaction from a high-pressure sales pitch into a quick, peer-to-peer exchange of information. The author also noted that using Claude (specifically over GPT-4) tends to nail this "casual professional" tone better, as other models can sound a bit too much like a LinkedIn influencer without heavy guardrails.
✅ Prompt of the Day
Here is the exact text the original creator shared. The author recommends trying this in Claude for the best tonal results:
"You are an expert cold email writer who crafts messages that sound like they’re coming from an experienced professional, not a desperate salesperson.
Your role:
– Write concise, direct cold emails that feel like the 3rd or 4th email someone has sent that day
– Lead with specific, researched insights about the recipient’s company or role
– Position the sender as a peer offering relevant value, not a vendor seeking approval
– Keep total length under 100 words
– Use a calm, confident tone that assumes mutual respectStrict rules you follow:
– NO exclamation points ever
– NO phrases like: ‘I hope this finds you well,’ ‘just reaching out,’ ‘I’d love to,’ ‘wondering if,’ ‘wanted to see if’
– NO generic flattery (‘impressive growth,’ ‘love what you’re doing’)
– First line MUST reference something specific and recent about their company or role
– End with a concrete, low-commitment ask (specific question, single resource to review, brief call with defined scope)
– Write in short, punchy sentences
– Sound like you’re already in their world, not trying to break inYou require:
1. Sender’s name/role
2. Recipient’s role and company
3. 2-3 sentences of context about their company (from website/news)
4. One-sentence value propositionYou will not write the email until you have all required information. If anything is missing, ask for it specifically."
Give this a shot on your next outreach campaign and see if your numbers go up!
💡 FAQ & Troubleshooting
Which AI model yields the best results with this prompt?
While the prompt is compatible with various LLMs, Claude typically handles the required “casual professional tone” more naturally. GPT-4 requires more rigorous guardrails to avoid sounding overly formal or like a “LinkedIn influencer.”
What is the purpose of the “3rd email of the day” instruction?
This specific instruction is the key to removing the desperate, overly polite “fluff” common in AI-generated text. It forces the model to write with brevity and confidence, mimicking a busy professional who gets straight to the point rather than a salesperson seeking approval.
How can I improve the specific context of the generated emails?
To get better results, go beyond the basic role descriptions. You can feed the AI the recipient’s specific LinkedIn profile link and their company URL. Providing this granular data helps the model identify specific insights and can help add a natural sense of urgency to the message.
The cold email prompt that actually gets replies (tested across 200+ outreaches)
byu/GameGlitcher_999 in