I Sent the Same Cold Email Brief to ChatGPT, Claude, and Gemini — Here's What Happened
Most people just pick one AI and stick with it out of habit. That's leaving money on the table. I gave all three major AI tools — ChatGPT, Claude, and Gemini — the exact same cold email brief, scored them on five criteria, and the results were not what I expected. Claude did something in its first draft that took me three rounds of prompting to get from ChatGPT. And Gemini? It surprised everyone, including me. By the end of this article, you'll know exactly which AI to use for cold emails depending on your goal — and you'll have the prompts to make any of them perform at their best.
The Brief I Used (And Why the Test Was Designed to Be Hard)
A generic brief gets you generic results. So I made this one specific enough to separate the average from the exceptional.
Here's the exact brief I gave all three: "Write a cold email to a VP of Marketing at a mid-size SaaS company. I'm selling an AI-powered analytics tool that cuts reporting time by 40%. Keep it under 150 words. No fluff. End with a soft CTA."
Same words. Same order. Pasted directly into ChatGPT-4o, Claude 3.5 Sonnet, and Gemini 1.5 Pro with zero additional context.
The five criteria I scored on: opening line quality, personalization depth, clarity of value proposition, tone, and CTA strength. Each scored out of 10. Maximum score: 50.
Here's what came back — and where each one broke down or broke through.
Round 1: ChatGPT Writes Safe. That's Both Its Strength and Its Problem.
ChatGPT's email was clean, professional, and totally forgettable.
It opened with: "Hi [Name], I hope this finds you well." That line alone should be illegal in cold email. To be fair, after I flagged it, one re-prompt fixed it immediately — ChatGPT is highly responsive to feedback, which is its real superpower.
The value prop was solid: it mentioned the 40% time reduction clearly and early. The CTA was polished: "Would you be open to a 15-minute call this week?" Safe, proven, fine.
Where ChatGPT lost points was originality of the opening and emotional resonance. It writes like a professional who's read every copywriting book but never felt the anxiety of actually sending a cold email. It's correct, not compelling.
Final score: 36/50. Best for: people who need a reliable first draft they can edit quickly. Use the prompt: "Rewrite the opening line of this cold email to hook a VP of Marketing who gets 200 emails a day — make it specific to their world."
Round 2: Claude Does Something Nobody Talks About — It Thinks About the Reader
Claude didn't just write an email. It wrote an email from the VP's perspective.
Without being asked, Claude's draft opened with: "Your team is probably spending Monday mornings rebuilding the same dashboards they built last Monday." That's a pain-point-first opener — and it landed immediately. It shows Claude isn't just filling in a template. It's modeling what the recipient actually experiences.
This is the insight most AI comparisons skip: Claude has the strongest theory of mind of the three. It thinks about what the other person is feeling before it thinks about what you want to say. For cold email, that's the difference between a response and a delete.
The value prop connected the 40% stat to a real emotion — getting Friday afternoon back — rather than just stating a number. The CTA was softer and more disarming: "Happy to send over a one-pager if that's easier than a call." That's smart. It lowers the barrier.
Where Claude stumbled slightly: it ran 162 words, nudging over the 150-word brief. It occasionally over-explains when it should trust the reader to connect dots. But the bones of that email were better than anything the other two produced on a first draft.
Final score: 44/50. Best for: cold emails where tone and empathy matter more than brevity. Your best move here is the prompt: "Write this cold email from the perspective of someone who deeply understands the recipient's daily frustrations — don't mention features until line three."
Round 3: Gemini Surprised Me With Structure — But Struggled With Soul
Gemini 1.5 Pro produced an email that was structurally perfect and emotionally flat.
The opening was better than ChatGPT's default — it went straight to the problem: "Reporting shouldn't eat your team's week." Short. Direct. Works. That's a legitimate hook.
But the middle of the email read like a product page. It listed benefits in a way that felt like it was checking boxes rather than having a conversation. There's a difference between writing an email and writing content about an email — Gemini occasionally crosses that line.
Where Gemini genuinely impressed me: it was the most consistent across multiple runs. I generated five versions. ChatGPT varied wildly in quality. Claude's ranged from brilliant to over-written. Gemini's five drafts were all within a narrow band of "pretty good." If you're building a cold email system at scale — think sequences, A/B testing, high volume — that consistency has real value.
The CTA was functional but generic: "Let's find 15 minutes." Not bad. Not memorable.
Final score: 38/50. Best for: teams running high-volume outreach who need reliable, consistent output. Sharpen it with: "Make the CTA feel more human and less like a scheduling bot. Give me three variations — one bold, one soft, one curious."
The Part Most People Get Wrong
Most people treat AI cold email as a one-prompt job. Write the prompt, copy the output, hit send. That's wrong — and it's why their reply rates are terrible.
The real workflow is prompt → score → re-prompt → refine. None of these three AIs produced a send-ready email on the first pass. Not even Claude. The difference is that Claude needed one refinement, ChatGPT needed two or three, and Gemini needed structural edits even when the words were fine.
The other mistake: using the same AI for every type of cold email. Selling a high-ticket consulting service to a CEO? Use Claude — the empathy and tone will do work. Running 500 outreach emails for a SaaS trial signup? Use Gemini for consistency, then layer in one ChatGPT-polished subject line per batch.
Stop asking "which AI is best." Start asking "which AI is best for this specific email, this specific recipient, and this specific goal." That question will make your outreach measurably better this week.
Key Takeaways
- Claude 3.5 Sonnet: Writes the most emotionally intelligent cold emails — best for high-stakes, low-volume outreach where one response matters.
- ChatGPT-4o: Most responsive to feedback and re-prompting — best when you want to iterate fast and control the final voice yourself.
- Gemini 1.5 Pro: Most consistent across multiple outputs — best for high-volume sequences where reliability beats brilliance.
- Pain-point-first openings: The single biggest difference between a cold email that gets read and one that gets deleted — lead with their problem, not your product.
- One-prompt drafts don't work: Every AI needs at least one refinement round — build that into your workflow and your reply rates will reflect it.
What to Do Right Now
Open Claude 3.5 Sonnet and paste this prompt: "Write a cold email under 150 words to a [job title] at a [company type]. I'm selling [product] that delivers [specific result]. Open with a line about a frustration they feel every week. End with a CTA that lowers commitment." Run the same prompt in ChatGPT immediately after, compare the opening lines side by side, and you'll feel the difference in about 90 seconds flat.