We Gave ChatGPT, Claude, and Gemini the Same Cold Email Brief — Here's What Happened
Most people just open ChatGPT and start typing. That's fine — until you realize you've been leaving serious quality on the table. We ran a head-to-head test, giving all three major AI writing tools the exact same cold email brief for a B2B SaaS pitch, and the results were genuinely surprising. Claude didn't just win — it won in a way that exposes a real weakness in how most people think about AI-generated copy. By the end of this article, you'll know exactly which tool to use for cold outreach, how to prompt it properly, and why the default choice is costing you replies.
We Gave All Three the Same Brief — Here's What They Produced
The brief was simple and realistic. A growth consultant wants to email a Head of Marketing at a mid-size e-commerce brand. The goal is to book a 20-minute call. The consultant has helped three similar brands increase email revenue by 30% in 90 days.
The exact prompt we used:
"Write a cold email from a growth consultant to a Head of Marketing at a mid-size e-commerce brand. The goal is to book a 20-minute call. The consultant has helped 3 similar brands increase email revenue by 30% in 90 days. Keep it under 150 words. Subject line included."
Same prompt. Three different tools. Here's the honest breakdown.
ChatGPT (GPT-4o) produced a clean, professional email. The structure was solid — hook, credibility, CTA. But it felt like every cold email you've ever ignored. The subject line was "Boost Your Email Revenue by 30%" — technically accurate, instantly deletable.
Gemini (1.5 Pro) surprised in one area and disappointed in another. It added a thoughtful opening line referencing "brands at your stage of growth," which showed some awareness of the audience. But the body got bloated fast — it hit 180 words and buried the CTA under two paragraphs of explanation. Gemini tends to over-explain when it should be cutting.
Claude (3.5 Sonnet) wrote the tightest, most human-sounding email of the three. The subject line was "Quick question about your email channel" — low pressure, high curiosity. The opening didn't start with "I." The CTA was a single, frictionless question. It read like a real person wrote it at 9am on a Tuesday.
Why Claude Wins at Cold Email (And What It's Actually Doing Differently)
Claude's edge isn't magic — it's a different default writing philosophy. Where ChatGPT optimizes for completeness and Gemini for thoroughness, Claude optimizes for conversational clarity. That's exactly what cold email requires.
Cold email lives and dies on one thing: does it feel like a human wrote this specifically for me? Claude's training makes it naturally resist filler phrases, passive voice, and the "professional tone" trap that makes most AI copy sound like a LinkedIn post from 2019.
Here's the specific thing Claude does that the others don't: it avoids the "I" opener by default. Starting a cold email with "I" is one of the fastest ways to signal it's about you, not them. Claude opened with "Your email channel is sitting on untapped revenue — and I've seen exactly this pattern with three other DTC brands." That's a reader-first construction. ChatGPT opened with "I help e-commerce brands grow their email revenue." One of those makes you keep reading. The other doesn't.
Claude also handles length discipline better than either competitor. It instinctively understood that 150 words means 150 words — not 150 words plus a P.S., not 148 words with a three-line disclaimer. Respecting the constraint is part of the craft.
This matters beyond just one test. If you're sending cold outreach at any volume — even 10 emails a week — the compounding difference in reply rate between "sounds like a robot" and "sounds like a thoughtful human" is massive. Even a 3% vs 8% reply rate difference means nearly 3x the conversations from the same effort.
How to Prompt Claude Specifically for Cold Email (The Right Way)
Knowing Claude is better for cold email is step one. Knowing how to prompt it properly is where most people still lose.
Start with a role + constraint + outcome structure. Don't just paste your brief. Frame it like this:
"You're a direct response copywriter who specializes in cold outreach. Write a cold email under 120 words with a subject line. The sender is [X], the recipient is [Y role] at [Z type of company]. The goal is [specific outcome]. Key proof point: [one specific result]. Rules: No 'I' opener. No buzzwords. End with one low-friction question."
The "Rules" section at the end is the single biggest upgrade you can make to any cold email prompt. Claude follows constraints more reliably than ChatGPT when they're explicit. If you don't say "no buzzwords," you'll get "synergize" somewhere in paragraph two.
Next, use Claude's editing mode after the first draft. Prompt it with: "Make this sound 20% more like a real person sent it and 20% less like a sales email. Don't change the structure." This second pass is where Claude really separates itself — it'll strip passive voice and replace corporate phrases without blowing up what's already working.
One workflow that takes under 10 minutes:
- Generate the first draft with the structured prompt above
- Ask Claude to give you 3 alternative subject lines — pick the one with the most curiosity, not the most clarity
- Run the "20% more human" edit pass
- Copy into your email tool
You'll have a cold email that's tighter, more human, and more likely to get a reply than anything you'd produce in a 30-minute manual writing session.
The Part Most People Get Wrong
Most people treat AI cold email like a copy-paste machine. They generate the email, skim it, hit send. That's wrong, and here's exactly why: AI writes for the average recipient, and your recipient isn't average.
The email Claude wrote in our test was genuinely good — but it didn't know anything about the specific person receiving it. It didn't know their company had just launched a new product line. It didn't know they posted about email deliverability struggles last week. One line of real personalization beats three paragraphs of perfect AI copy every time.
The right workflow isn't "use AI to write the whole email." It's "use AI to write 90% of the email, then spend 60 seconds adding one specific, human detail." That detail is what makes the recipient think "wait, this person actually looked at my stuff."
The second mistake: people default to ChatGPT because it's the one they know. That's fine for most tasks. But for copy that needs to sound human and concise — cold email, DMs, LinkedIn outreach — Claude's defaults are genuinely better-calibrated. Using the right tool for the right job isn't brand loyalty, it's just smart.
Key Takeaways
- Claude 3.5 Sonnet: Produces the most human-sounding cold email of the three tools tested — especially for concise, reader-first copy
- ChatGPT (GPT-4o): Reliable and structured, but defaults to corporate tone that feels templated in sales contexts
- Gemini 1.5 Pro: Shows audience awareness but over-explains — best used for research and drafting, not final outreach copy
- The "Rules" prompt section: Adding explicit constraints (no "I" opener, no buzzwords, end with a question) is the single biggest quality upgrade you can make to any AI email prompt
- The 90/10 rule: Let AI write 90% of the cold email, then spend 60 seconds adding one specific, researched detail about the recipient — that last 10% is what actually gets replies
What to Do Right Now
Open Claude at claude.ai and paste this prompt: "You're a direct response copywriter. Write a cold email under 120 words with a subject line. The sender is [your name/role], the recipient is [target role] at [target company type]. Goal: book a 20-minute call. Key proof point: [your best result]. Rules: No 'I' opener. No buzzwords. End with one low-friction question." Fill in your details, run the draft, then do one follow-up prompt: "Make this 20% more human and 20% less salesy." You'll have a genuinely strong cold email in under 10 minutes — then compare it against what ChatGPT produces with the same brief and see the difference yourself.