We Fed the Same Terrible Sales Email to ChatGPT, Claude, and Gemini — Here's What Happened
We took a genuinely bad sales email — the kind that starts with "I hope this email finds you well" and ends with "please let me know if you have any questions" — and ran it through all three major AI tools. The results were so different they almost feel like they came from different planets. One AI rewrote it like a seasoned copywriter. One played it safe and gave us something forgettable. And one produced an output so generic it would've gone straight to the trash folder. By the end of this article, you'll know exactly which AI to trust with your sales copy, why the "best" AI isn't always obvious, and how to write a prompt that gets you results you can actually send. Let's start with what we were actually working with.
The Original Email Was Genuinely Painful to Read
Here's the email we used as our test case. No changes, no cleanup — exactly the kind of thing that lands in inboxes every single day:
"Hi [Name], I hope this email finds you well. I wanted to reach out because I think our software could really help your business. We have a lot of great features and many clients have seen great results. Would love to set up a 15-minute call to chat more. Please let me know if you have any questions. Best, Dave"
This email has every fatal sin of B2B outreach. It's vague, it's self-focused, it has no hook, no specific value proposition, and the call-to-action is so weak it practically apologizes for existing.
We gave the same prompt to all three AIs: "Rewrite this sales email to make it more compelling, specific, and likely to get a response. Keep it under 150 words."
No extra context. No role-playing instructions. No persona setup. Just the raw prompt — because that's how most people actually use these tools, and that's the honest test.
The word limit was intentional. Great sales emails are short. If an AI can't tighten your copy under pressure, it won't help you in the real world.
What Each AI Actually Produced — And Why the Differences Matter
ChatGPT (GPT-4o) came out swinging. It immediately dropped the "I hope this email finds you well" opener and replaced it with a direct, problem-focused line: "Most [industry] teams lose 6+ hours a week on [pain point] — and don't realize there's a fix." It invented a specific pain point placeholder, added social proof framing ("We helped [Company X] cut that in half in 30 days"), and closed with a low-friction CTA: "Worth a 10-minute look?" The result felt like something a real sales rep would actually send.
Claude (Claude 3.5 Sonnet) took a different approach entirely. Instead of leading with pain, it opened with curiosity: "Quick question — how much time does your team spend on [specific task] each week?" This is a classic consultative selling move, and it works. Claude's rewrite was tighter, more conversational, and had a natural rhythm that felt human. It also added a subtle credibility marker without being braggy. Honestly, Claude's version was the most immediately usable out of the box.
Gemini (Gemini 1.5 Pro) disappointed. Its rewrite was polished on the surface — grammatically clean, professionally structured — but it kept the same skeleton as the original. It still opened with a soft intro, still buried the value prop in the second paragraph, and still used "I'd love to connect" as the CTA. It dressed up the email without fixing it. That's a critical difference most people miss.
The gap between ChatGPT and Claude was small but real. The gap between those two and Gemini was significant enough to matter in a real campaign.
This tells you something important: the AI you use for writing tasks is not a neutral choice. The same weak prompt produces wildly different results depending on where you type it.
The Hidden Variable Nobody Talks About: Prompt Quality Changes Everything
Here's what the basic test doesn't show you. When we upgraded the prompt, the rankings shifted.
We ran a second round with a role + context + constraint prompt instead of a bare instruction. The new prompt was: "You are an expert B2B sales copywriter. Rewrite this email for a SaaS company targeting operations managers at mid-sized logistics firms. The goal is to get a reply, not a meeting. Keep it under 120 words. Use a direct, confident tone — no fluff."
With that prompt, Gemini improved dramatically. It started leading with specificity, trimmed the fat, and produced something actually worth testing. Claude got even sharper — almost surgical in how it targeted the persona. ChatGPT stayed strong but leaned slightly more formal than necessary.
The lesson: Gemini needs more context to shine. It's not a bad tool — it's a context-hungry tool. Give it vague instructions and it produces vague output. Give it a detailed brief and it punches closer to its weight.
This is the insight most "AI comparison" articles skip entirely. They test all tools with the same generic prompt and call it a day. But in the real world, you're going to refine your prompts — and knowing how each AI responds to context is more valuable than knowing which one wins on round one.
Think of it like this: ChatGPT is a good writer who works well with minimal direction. Claude is a thoughtful editor who asks smart clarifying questions (sometimes literally — it'll ask what you mean if your prompt is unclear). Gemini is a capable analyst who needs a proper brief before it produces its best work.
The mental model to take away: Match the AI to how much context you have to give. Tight deadline, rough prompt? Use ChatGPT or Claude. Time to write a proper brief? Gemini becomes more competitive.
How to Use This Today: A 15-Minute Sales Email Upgrade
Pick one sales email in your drafts right now — the one you've been stalling on sending because it doesn't feel right. That's your test subject.
Step 1: Run it through Claude first. Use this prompt: "You are a direct-response copywriter. Rewrite this sales email to lead with a specific problem the reader likely has, show you can solve it in one sentence, and close with one low-friction question. Under 100 words. No filler phrases." Claude's conversational instinct makes it the best first pass for cold outreach.
Step 2: Take Claude's output into ChatGPT for a punch-up. Prompt: "Make this email punchier. Cut any sentence that doesn't pull the reader forward. The first line needs to stop a scroll." ChatGPT's headline instincts are strong — it'll find the weak opener Claude missed.
Step 3: Use Gemini for a reality check. Paste both versions and ask: "Which of these two emails is more likely to get a reply from a busy operations manager who gets 80 emails a day? Explain your reasoning in 3 bullet points." Gemini's analytical framing makes it genuinely useful for evaluation, even if it struggled at generation.
You now have three drafts, a critique, and a clear winner — in under 15 minutes. A/B test the top two. Check your reply rates after 20 sends. Iterate from there.
This workflow turns AI tools from novelties into an actual competitive advantage in your outreach.
The Part Most People Get Wrong
Most people use AI to write their sales emails from scratch. That's the wrong move. AI-generated cold emails written from zero almost always sound like AI-generated cold emails — and your prospects can tell.
The right use of these tools is rewriting and sharpening, not creating from nothing. You write the core message — the real insight about your prospect, the genuine value you offer — and then you use AI to tighten the language, cut the dead weight, and sharpen the hook. Your authenticity, their efficiency.
The other mistake: people run one version, get something decent, and stop there. Every AI has a different editorial instinct. Running your email through all three takes 10 minutes and gives you three completely different angles to evaluate. You're not just getting a better email — you're learning what a better email actually looks like.
And finally: don't ignore the tool that "lost." Gemini's evaluation mode — asking it to critique rather than create — is underused and genuinely powerful. Even in our test, its best output wasn't a rewrite. It was a clear-eyed analysis of why the original email failed. Know your tools' strengths.
Key Takeaways
- ChatGPT: Best for punchy, hook-driven rewrites with minimal prompting — strong editorial instincts out of the box.
- Claude: Most consistently human-sounding output for sales copy, especially when you need a conversational, curiosity-driven tone.
- Gemini: Needs a detailed brief to perform at its best — but excels at evaluation and analytical critique when creation isn't the goal.
- Prompt quality: The gap between tools shrinks dramatically when you give them a role, context, and clear constraints — don't skip this.
- The right workflow: Use AI to rewrite and sharpen your copy, not generate it from nothing — your authentic insight plus their editorial power is the winning combo.
What to Do Right Now
Open Claude, paste your worst-performing sales email, and use this prompt: "You are a direct-response copywriter. Rewrite this email to lead with a specific pain point, prove you can solve it in one sentence, and close with a single low-friction question. Under 100 words. No filler." Then take that output to ChatGPT and ask it to make the first line impossible to ignore. You'll have a send-ready email in under 10 minutes — and you'll never look at your old template the same way again.