ChatGPT vs Claude vs Gemini: Rewriting a Weak Sales Email
I Gave All Three the Same Terrible Sales Email — Only One Made Me Want to Buy
Most AI comparison articles test these tools on trivia questions or coding puzzles. That's not how most people actually use AI. I wanted to test something real — the kind of task that directly affects whether someone makes money or doesn't. So I took a genuinely bad sales email (the kind that sits in your inbox making you feel nothing), fed it to ChatGPT, Claude, and Gemini, and asked each one to fix it. The results were not even close. One version felt like it was written by a copywriter who'd actually sold something before. The other two? Nice, but forgettable. Here's exactly what each tool did, where it failed, and what you should steal from the winner.
The Original Email Was Bad on Purpose — Here's What "Weak" Actually Looks Like
Before I show you the rewrites, you need to see what I started with. Because a lot of sales emails people write — or get — look exactly like this.
Here's the original:
"Hi there, I hope this email finds you well. I wanted to reach out because I think our product could be a great fit for your business. We offer a range of solutions that help companies improve efficiency and grow revenue. Let me know if you'd like to schedule a call to learn more. Best regards, Dave."
That email has every classic problem. No specific pain point. No real hook. No urgency. "I hope this email finds you well" is the sales equivalent of a limp handshake. "A range of solutions" tells the reader exactly nothing.
The prompt I gave each AI was identical: "Rewrite this sales email to be more persuasive. Make it specific, compelling, and focused on the reader's benefit. Keep it under 150 words."
I kept the prompt simple on purpose. The whole point was to see what each AI naturally does when given a vague brief — because that's what most people send.
Every AI had the same input, the same instruction, and the same word limit. What came back told me a lot about how each one thinks.
What ChatGPT, Claude, and Gemini Actually Produced — Side by Side
ChatGPT (GPT-4o) went for clean and professional. It cut the fluff, tightened the structure, and added a clearer call to action. The rewrite opened with: "Most [industry] teams waste 6+ hours a week on tasks your competitors have already automated." Solid. It used a specific number to create curiosity and grounded the email in a real problem.
But here's the issue — the specificity felt performed. ChatGPT inserted a placeholder stat (the "6+ hours") that it invented. It wasn't grounded in anything real. The structure was great, the instinct was right, but it felt like a template that'd been filled in, not a message written for a human.
Gemini (Google's model) gave me the longest rewrite — and the weakest. It kept phrases like "we'd love the opportunity to connect" and buried the value proposition in the third paragraph. Gemini seemed to optimize for politeness over persuasion. It's a solid tool for a lot of tasks, but on copywriting instinct? It plays it too safe.
Claude (Claude 3.5 Sonnet) surprised me. Instead of just restructuring the email, it opened with a direct question: "Are you still doing manually what your competitors automated six months ago?" Then it named a specific outcome, created a soft deadline, and closed with a one-line call to action that didn't feel pushy. The whole thing read like it was written to someone, not at them.
Claude understood something the other two missed: great sales copy doesn't sell the product, it sells the reader's desired future. That's the difference between a rewrite that sounds better and one that actually converts.
The Hidden Skill That Separates Good AI Copy from Great AI Copy
Here's what most comparison articles won't tell you: the rewrite quality has less to do with which AI you use and more to do with how much context you give it.
When I gave Claude a second, richer prompt — "Rewrite this sales email. The reader is a busy operations manager at a 50-person company. They're frustrated with manual reporting. Our tool automates that in under 10 minutes. Be direct, empathetic, and end with one clear CTA." — the output jumped from good to genuinely impressive.
This is the mental model you need: AI rewrites to the quality of your brief. The tool isn't the ceiling — your context is. Think of it like hiring a freelance copywriter. If you hand them one bad email and say "fix it," you'll get something decent. If you hand them a buyer persona, a pain point, and a desired outcome, you'll get something that works.
The technique is called role + pain + outcome + constraint briefing. You tell the AI who the reader is, what's frustrating them, what success looks like, and what limits to work within. Every single time I added this structure to my prompts, the rewrites got sharper — across all three tools.
ChatGPT responds really well to explicit formatting instructions alongside context. Claude thrives when you describe the emotional state of the reader. Gemini improves most when you tell it what not to do ("don't use corporate jargon, don't open with pleasantries").
That's the insight most people skip. They compare the tools at their baseline. You should compare them at their ceiling — which means prompting them properly first.
How to Run This Exact Test on Your Own Sales Emails Today
You don't need to take my word for any of this. Here's how to replicate this in the next 20 minutes and get something you can actually send.
Step 1: Grab your worst-performing email. Find one with a low open rate, a low reply rate, or one you've been embarrassed to send but sent anyway. Copy the full text.
Step 2: Build your brief using the role + pain + outcome + constraint formula. It looks like this: "Rewrite this sales email. The reader is a [job title] at a [company size] company who struggles with [specific frustration]. Our [product] helps them [specific outcome] in [timeframe]. Be direct and human. No corporate language. Under 150 words. End with one clear CTA."
Step 3: Run it through Claude first. Paste both your original email and the brief together. Claude 3.5 Sonnet is free on claude.ai — no subscription needed for basic use. Read the output out loud. If it sounds like something a real human would write, it's working.
Step 4: Run the same brief through ChatGPT. Use GPT-4o if you have Plus, or GPT-3.5 for free. Compare the two side by side. You'll often find Claude is stronger on emotional tone, while ChatGPT is stronger on structure and clarity. The best move is to combine them — take Claude's opening line, ChatGPT's structure, then edit the whole thing yourself for 5 minutes.
Step 5: Test Gemini last. It'll often give you a safer, more polished version — useful as a "would this embarrass me?" check, less useful as your primary rewrite engine. Think of Gemini as your final proofreader, not your first drafter.
That's the workflow. Three tools, one brief, 20 minutes, one email that's actually worth sending.
The Part Most People Get Wrong
Most people use AI to polish weak emails and call it done. That's wrong. Polishing a weak email gives you a shiny weak email.
The real problem is almost never the writing — it's the strategy behind the writing. If your email doesn't have a clear single reader, a specific pain point, and one obvious next step, no AI on earth will save it. It'll just sound more confident while saying nothing.
Before you ask any AI to rewrite anything, spend 2 minutes answering three questions on paper: Who exactly is reading this? What are they frustrated about right now? What's the one thing I want them to do after reading? If you can't answer all three, fix that first. Then bring in the AI.
The other mistake? Accepting the first draft. Every AI rewrite is a starting point, not a finished product. The people getting real results from these tools treat the first output as raw material — they cut lines that sound like AI, punch up the opener, and make sure the CTA is specific ("Book a 15-minute call here" beats "Let me know if you'd like to chat" every single time).
Key Takeaways
- Claude 3.5 Sonnet: Strongest at emotional tone and reader empathy — use it as your primary rewrite engine for sales copy.
- ChatGPT (GPT-4o): Excellent at structure and clarity — best used to tighten and organize after Claude gives you the raw draft.
- Gemini: Plays it safe and stays polished — most useful as a final check, not your lead copywriting tool.
- Role + Pain + Outcome + Constraint: The four-part prompt formula that raises AI rewrite quality across every tool — don't skip it.
- First draft = raw material: The biggest mistake is treating AI output as finished copy — edit it, read it aloud, and make it sound like a human wrote it.
What to Do Right Now
Open claude.ai, paste your weakest sales email, and add this brief above it: "Rewrite this. The reader is [job title] who struggles with [pain point]. My product helps them [outcome]. Be direct, human, under 150 words, one CTA." Read the result out loud. If you wouldn't say it to someone's face, edit it until you would. That's your new email.