I Gave ChatGPT, Claude, and Gemini the Same Terrible Sales Email — Here's What Happened
I gave all three major AI tools the exact same weak, cringe-worthy sales email and asked them to rewrite it. The results were wildly different — one AI turned it into something you'd actually want to send, one played it safe, and one somehow made the original look better. If you're using AI to write sales copy right now without knowing which tool actually wins at persuasion, you're leaving serious money on the table. By the end of this article, you'll know exactly which AI to open, which prompt to use, and why the "obvious" choice isn't always the right one.
The Terrible Email I Used (And Why It's the Perfect Test)
Here's the original email I fed into all three AIs, word for word:
"Hi there, I wanted to reach out because we have a great product that I think could really help your business. We've been in the industry for 10 years and have helped many clients. Our solution is cost-effective and easy to use. Please let me know if you'd like to learn more. Thanks."
This email is a disaster in five different ways — no personalization, no specific pain point, no social proof with real numbers, no clear offer, and a call-to-action so weak it practically apologizes for existing. It reads like it was written by someone who didn't want to be rude rather than someone who wanted to sell something.
I picked this email on purpose. Vague, passive, corporate fluff is the most common type of bad sales email in the wild. If an AI can fix this, it can fix yours.
The prompt I gave each AI was identical: "Rewrite this sales email to be more persuasive and compelling for a B2B audience. Keep it under 150 words." Same instruction, same word limit, same bad raw material. The only variable was the AI itself.
What Each AI Actually Produced — Ranked From Worst to Best
Let's start with the one that surprised me the most in the wrong direction: Gemini.
Gemini's rewrite was cleaner than the original, sure. It tightened the grammar and swapped out a few weak phrases. But it kept almost all the structural problems intact — still no specific pain point, still a soft CTA, still no real reason for the reader to care. The email went from a 3/10 to maybe a 4/10. It polished the hood of a broken car without looking under the engine.
ChatGPT (GPT-4o) did what it always does — produced something competent, structured, and readable. It added a semi-specific opener ("I noticed your team is scaling fast..."), dropped in a generic stat ("our clients see an average 30% increase in efficiency"), and ended with a clearer CTA. It's a solid B-minus rewrite. You could send it and not embarrass yourself. But it felt like a template, not a conversation.
Claude (claude-3-5-sonnet) is where things got genuinely interesting. Without any extra coaching, Claude rewrote the email with a different philosophy — it led with a problem statement, used specificity to create credibility, and wrote a CTA that felt like an invitation rather than a request. Here's a close approximation of what it produced:
"Subject: Quick question about [Company]'s Q4 pipeline Hey [Name], most [industry] teams I talk to are hitting the same wall right now — too many leads, not enough time to qualify them properly. We've helped companies like [Similar Company] cut their qualification time by 40% without adding headcount. Worth a 15-minute call this week to see if we'd be a fit? No deck, no pressure — just a real conversation."
That's not just "better." That's a fundamentally different approach to persuasion. Claude didn't just edit the email — it understood why the original failed.
The Hidden Difference Nobody Talks About: Intent vs. Instruction
Here's what most AI comparison articles completely miss: the gap between these tools isn't about vocabulary or grammar. It's about whether the AI understands persuasive intent or just follows surface-level instructions.
When you say "make this more persuasive," a shallow AI hears "use stronger adjectives." A deeper AI hears "find the emotional and logical levers that make a reader take action, then restructure the whole thing around those."
Claude consistently performs better on persuasion tasks because it models the reader's perspective, not just the writer's request. It thinks about who's going to open this email, what they're probably already dealing with, and what would make them pause. That's a different cognitive process than "improve the writing."
This is why prompt engineering matters so much here. If you give Claude a richer brief, the output gets dramatically better. Try this prompt: "Rewrite this sales email. The reader is a VP of Sales at a 50-person SaaS company. They're skeptical of cold outreach, short on time, and their Q4 target is at risk. Write from the angle that we understand their pressure, not that we have a great product."
That single context shift produced an email in my test that I would genuinely use. The more specific your situation, the more Claude's reasoning capability compounds.
ChatGPT responds better to structural prompts than emotional ones. Tell it "rewrite this using the AIDA framework" or "add a social proof hook in the first line" and it executes precisely. It's a strong tool when you already know the persuasion framework and just need it applied.
The mental model to carry: Claude is your strategist. ChatGPT is your executor. Gemini, at least for this kind of nuanced copy task, is still catching up.
How to Run This Test Yourself in the Next 15 Minutes
Take your worst-performing sales email — the one with the lowest reply rate, the one you're a little embarrassed by — and run it through this exact workflow.
Step 1: Open claude.ai and paste this prompt: "Here's a sales email I've been sending. Before you rewrite it, tell me the three biggest reasons it's probably not converting. Then rewrite it addressing those reasons." Paste your email below the prompt. Read the diagnosis carefully — that alone is worth more than most copywriting courses.
Step 2: Take Claude's rewrite and open ChatGPT (GPT-4o). Paste this prompt: "Here's a sales email draft. Rewrite it in three different tones: confident and direct, warm and consultative, curiosity-driven. Keep each version under 120 words." Now you have three variants to A/B test.
Step 3: Pick your strongest variant and run it through Gemini with this prompt: "Check this email for any phrases that sound like generic marketing copy or AI-written text. Suggest more human, natural replacements." Gemini is actually solid at this cleanup task — it's where it earns its place in the workflow.
This three-tool workflow takes about 15 minutes and gives you a tested, human-sounding email that's been pressure-checked from multiple angles. Most people use one AI once. Running all three in sequence is how you get results that actually feel different in your inbox stats.
The Part Most People Get Wrong
Most people treat AI like a magic "make it better" button — they paste in their bad email, hit generate, and send whatever comes out. That's wrong, and here's why: the AI can only work with the information you give it.
If you don't tell the AI who's reading the email, what problem they have, what makes your offer different, and what you want them to do — the AI will fill in those blanks with the most generic possible assumptions. That's how you end up with output that's grammatically clean and emotionally empty.
The fix is a two-step process. First, brief the AI like you'd brief a copywriter. Describe the reader, the context, the goal, and any specific objections. Second, ask the AI to explain its choices. Prompt: "Tell me why you structured the email this way and what persuasion principle you used." If it can't explain the logic, the copy probably doesn't have any.
The other mistake is using Gemini for heavy persuasion lifting right now. It's genuinely improving — and it's excellent for research, summarization, and Google Workspace integration. But in this specific test, on this specific task, it underdelivered. Use the right tool for the right job.
Key Takeaways
- Claude wins on persuasion: For emotional, strategic, reader-centric copy tasks, Claude outperforms the others without extra prompting.
- ChatGPT wins on execution: Give it a framework (AIDA, PAS, before/after/bridge) and it applies it precisely — great when you know what you want.
- Gemini is best for cleanup: Use it at the end of your workflow to strip out generic phrases and polish for human tone, not as your primary rewrite tool.
- Context is everything: The more specific your prompt brief (reader, pain point, goal, objection), the better every AI performs — but Claude benefits most from this detail.
- The three-tool workflow beats any single tool: Diagnose with Claude, generate variants with ChatGPT, humanize with Gemini — 15 minutes, dramatically better output.
What to Do Right Now
Open claude.ai, paste your weakest sales email, and use this exact prompt: "Before rewriting this, give me the three biggest reasons it's probably not converting. Then rewrite it for a [your target reader, e.g., 'skeptical VP of Marketing at a mid-size e-commerce company'] who values specificity over hype." Read the diagnosis first — it'll change how you write every email after this.