I Gave ChatGPT, Claude, and Gemini the Same Cold Email Brief — One Clear Winner Closed More Deals
Most AI tools can write a cold email. But "write a cold email" is like saying a car "gets you places" — the difference between a beat-up Honda and a BMW matters a lot when you're trying to close a $10,000 deal. I ran a head-to-head test: the exact same brief, the exact same prospect persona, fed into ChatGPT, Claude, and Gemini. The outputs were dramatically different in tone, structure, and — most importantly — real-world reply rate. One tool understood persuasion at a level the others simply didn't. Here's exactly what happened, what I learned, and how you can use it to write cold emails that actually get responses starting today.
The Test: Same Brief, Three AI Tools, Three Very Different Emails
The brief was simple and realistic. I asked each tool to write a cold email to a SaaS founder with 10–50 employees, pitching a fractional CFO service. The prospect had recently raised a seed round and was likely burning cash without a financial strategy. The goal: get a 20-minute discovery call booked.
The prompt I used was identical across all three tools:
"Write a cold email to a SaaS founder who just raised a seed round. I'm pitching fractional CFO services. The goal is to get a 20-minute discovery call. Keep it under 150 words. Don't use buzzwords. Sound human."
ChatGPT (GPT-4o) came back fast — confident, clean, and structured. It opened with a compliment on the funding round, mentioned a pain point (burn rate), and ended with a soft CTA. It was good. Genuinely good. But it read like it was written by someone who had studied cold emails, not someone who had lived them. The phrasing was slightly too polished — "I help post-seed SaaS founders..." — which is the kind of opening that gets skimmed.
Gemini gave me something that felt like a press release wrote a cold email. It was longer than asked, hit four bullet points of service features, and ended with "I'd love to connect at your earliest convenience." That phrase alone would tank your open-to-reply rate. Gemini struggled most with the human-sounding constraint — it kept reverting to formal, corporate-adjacent language.
Claude (Claude 3.5 Sonnet) did something the other two didn't. It led with a specific, uncomfortable truth about post-seed SaaS companies — that most founders discover their runway math is wrong around month eight, not month two. Then it made the ask feel like an observation, not a pitch. The email felt like it came from a peer, not a vendor. That distinction is worth real money.
Why Claude's Output Worked: The Psychology It Got Right
The gap between Claude's email and the others wasn't about word choice — it was about psychological positioning. Claude instinctively wrote from the prospect's internal world rather than the sender's external pitch.
Most cold emails are structured as: Here's who I am → Here's what I do → Here's my CTA. That's a logical sequence, but it's the wrong emotional sequence. The reader doesn't care who you are yet. They care about their own problem. Claude's output flipped this to: Here's a problem you're probably feeling right now → Here's a moment it gets worse → Here's who can help.
This is called problem-first framing, and it's one of the most reliable conversion techniques in direct response copywriting. Claude seems to have absorbed enough sales psychology in its training that it applies this instinctively — even when you don't explicitly ask for it.
Here's the specific line Claude produced that made the difference: "Most founders I talk to don't realize their runway calculation is off until they're 4 months from zero." That sentence does three things at once — it establishes credibility, triggers loss aversion, and implies the sender has talked to people just like them. ChatGPT's opener was good. That line is great.
The lesson isn't "always use Claude." The lesson is to understand why that line worked, so you can prompt any tool to produce it. Which is exactly what we're going to do next.
How to Prompt Any AI Tool to Write Like Claude Did
You don't need to switch tools — you need to give better instructions. The difference between a mediocre cold email and a great one often comes down to what you tell the AI before it writes a single word.
Start with what I call the "Internal Monologue Prompt." Before asking for the email, tell the AI exactly what your prospect is worried about right now. Here's the structure:
"Before writing this email, describe what my prospect is privately stressed about this week. Then write the email from that emotional context."
When I ran this in ChatGPT, the output quality jumped dramatically — nearly matching Claude's instinctive approach. The AI stopped writing at the prospect and started writing for them. That single instruction change is worth testing today.
The second technique is contrast-based specificity. Vague cold emails die in the inbox. Instead of prompting "write a cold email," try:
"Write a cold email that opens with a specific, uncomfortable truth about [industry/role], followed by a single-sentence pitch, followed by a low-friction CTA that takes less than 20 minutes of their time."
Gemini improved the most from this constraint. When you remove its freedom to pad and bullet-point, it produces something far leaner. The constraint is the feature.
Finally, always add a "sound like" reference point to your prompt. Try: "Sound like a founder who's been in this person's exact position, not a vendor trying to sell them something." This reframes the AI's tone model completely. Across all three tools, this instruction consistently produced warmer, less formal output that reads as peer-to-peer rather than B2B pitch.
The Part Most People Get Wrong
Most people treat AI cold email tools like a vending machine — drop in a prompt, collect an email, hit send. That's the fastest way to build a reputation for sending forgettable, templated outreach.
The mistake is using the AI's first output as the final output. Every email these tools produce is a first draft, not a finished product. Your job is to take the best 1–2 sentences from what the AI gives you and rewrite the rest in your own voice. That hybrid approach — AI structure, human voice — consistently outperforms 100% AI copy or 100% manual copy.
Here's a dead-simple workflow: run your prompt in both ChatGPT and Claude, put the outputs side by side, and steal the strongest opening line from one and the strongest CTA from the other. Takes 5 minutes. Produces something neither tool created alone.
The second big mistake is ignoring subject line testing. The best email body in the world doesn't matter if nobody opens it. Ask each tool specifically: "Give me 5 subject lines for this email — make three curiosity-based and two direct." Then A/B test two. Most email tools (Instantly, Lemlist, Apollo) make this trivially easy. Most people skip it entirely.
Key Takeaways
- Claude 3.5 Sonnet: Produces the most psychologically sophisticated cold emails by default, especially for B2B pitches targeting founders or executives.
- Problem-first framing: Leading with the prospect's internal stress — not your pitch — is the single biggest lever for improving cold email reply rates.
- The Internal Monologue Prompt: Asking AI to describe your prospect's private stress before writing dramatically improves output quality across all three tools.
- First draft rule: Never send an AI's first output — use it as raw material, then rewrite in your own voice for the final 30% of the email.
- Subject line neglect: Most people optimize the email body and ignore the subject line, which determines whether any of this work even matters — always generate and test at least two options.
What to Do Right Now
Open Claude at claude.ai, paste in this prompt: "Before writing anything, describe what a SaaS founder who just raised a seed round is privately stressed about this week. Then write a cold email pitching [your service] in under 130 words — lead with their pain, not your pitch, and end with a CTA that takes less than 20 minutes." Replace the bracket with your actual service, read what comes back, and rewrite the opening line in your own words. That one email, sent to 10 targeted prospects this week, will teach you more about AI-assisted outreach than any other article you'll read today.