I Let AI Write 500 Cold Emails. Here's What Actually Got Replies.
The Setup
We took 500 real prospects — small business owners in trades, fitness, and e-commerce — and sent each one a cold email. Five groups of 100, five different AI-generated approaches. Same offer, same sender, same time of day. Only the email strategy changed.
Here's what we tested:
Approach 1: The "I Noticed" Email
"Hey {name}, I noticed your Google reviews dropped from 4.8 to 4.3 this quarter..."
Classic consultative cold email. AI researched each business, found a specific data point, and opened with it. Sounds smart. Sounds like what every sales guru tells you to do.
Reply rate: 3.2%
Decent. Not great. Turns out when everyone does the "I noticed" opener, nobody notices yours.
Approach 2: The Blunt Offer
"We find leads for {industry} businesses. 20 scored contacts delivered to your inbox weekly. Free trial, no calls needed."
No personalization. No flattery. Just the offer.
Reply rate: 7.1%
More than double the "thoughtful" approach. Huh.
Approach 3: The Question Hook
"Quick question — if someone handed you 15 qualified leads every Monday morning, what would that be worth to your {business_type}?"
One question, one line, one curiosity gap.
Reply rate: 8.4%
Winner winner. Short, punchy, makes them do math in their head.
Approach 4: The Social Proof Stack
"We just delivered 47 leads to a {similar_business} in {nearby_city}. Their owner said he's booked through August."
Specific proof, geographic proximity, implied scarcity.
Reply rate: 6.8%
Strong. The specificity of "47 leads" and a real city made it feel less like spam.
Approach 5: The Anti-Sales Email
"This is a sales email. I'm not going to pretend it isn't. Here's what we do: [one sentence]. Here's the catch: [one sentence]. Want in?"
Radical honesty. Meta-awareness. AI generated the one-liners per business type.
Reply rate: 5.9%
Respectable. Some people loved the honesty. Others ignored it.
The Uncomfortable Truth
The hyper-personalized, research-heavy email that sales Twitter worships? Dead last. The short, blunt, value-first emails crushed it.
Here's why: small business owners get 200 emails a day. They don't care that you "noticed" their website. They care about one thing — "will this make me money?"
The winning formula was embarrassingly simple: short + specific + a question that forces mental math.
AI's job wasn't to sound human. It was to be so clear and direct that the prospect couldn't ignore the value proposition.