I Gave the Same Awful Sales Email to 3 AIs. The Winner Surprised Me — and the Loser Wasn't Who You'd Expect.
Most people assume the biggest, most hyped AI wins every task. That assumption just cost a lot of salespeople real deals. I took one genuinely terrible cold sales email — the kind with three exclamation points, a "circle back," and zero reason for the recipient to care — and fed it word-for-word to ChatGPT, Claude, and Gemini. The results were not what I expected, and after you see them side by side, you'll never use the wrong AI for writing tasks again. Let me show you exactly what happened.
The Awful Email I Used (And Why I Chose It)
The email I used wasn't a strawman. It's the kind of email that lands in inboxes every single day, written by real people with real products who just don't know how to write for sales.
Here it is, unedited:
"Hi [Name], hope you're doing well! I wanted to reach out because we have an amazing solution that could really help your business. Our platform has helped many companies increase their ROI and streamline their operations. I'd love to set up a quick call to tell you more. Let me know if you're free this week. Looking forward to connecting!"
It's vague. It's self-centered. It has no hook, no specific pain point, no social proof with numbers, and no clear ask. The subject line was blank — I left that intentional so I could see what each AI would invent.
This email is a perfect test because fixing it requires more than just cleaning up grammar. A great rewrite needs to add a hook, create specificity out of thin air, flip the focus to the reader, and still sound human. That's a real cognitive task — not just autocomplete.
I gave all three AIs the exact same prompt: "Rewrite this cold sales email to make it more compelling, specific, and likely to get a response. Add a subject line. Keep it under 150 words."
What ChatGPT, Claude, and Gemini Actually Produced
ChatGPT (GPT-4o) rewrote the email fast and confidently. It invented a specific pain point ("struggling to consolidate your sales pipeline"), added a credible-sounding stat ("teams like yours cut reporting time by 40%"), and wrote a subject line that didn't suck: "Cut your reporting time by 40% — worth 15 minutes?" The tone was punchy, the ask was clear, and it came in at 143 words. It read like something a competent sales consultant would write.
Claude (Claude 3.5 Sonnet) did something different. Instead of just rewriting, it asked a clarifying question first — "Do you want me to target a specific industry or role?" — before producing a version. When I told it to just proceed with assumptions, it wrote the most structurally sophisticated version of the three. Claude identified that the original email was self-centered and explicitly flipped the frame, opening with the recipient's likely problem instead of the sender's product. The subject line: "The reason your reps are losing deals on follow-up." That's genuinely good.
Gemini (Gemini 1.5 Pro) gave me the most polished-looking output at first glance — clean formatting, a subject line, a clear CTA. But when I read it closely, something felt off. It was still vague in places. Phrases like "optimized workflows" crept back in. It looked like a rewrite but didn't fully solve the original problem. It swapped some words but kept the same empty energy.
Here's the ranking in order of effectiveness: Claude → ChatGPT → Gemini. Claude won on strategy. ChatGPT won on speed and reliability. Gemini looked good but underdelivered on the actual brief.
The loser most people wouldn't expect? Gemini — Google's AI, with all that data, all that infrastructure, underperformed on a simple but precise writing task. That matters.
The Hidden Difference: How Each AI Thinks About Writing
This is the part most comparison articles skip entirely. The outputs are different because the approach is different — and understanding that changes how you use these tools.
ChatGPT is a confident executor. You give it a task, it does the task. It makes assumptions, fills gaps, and delivers something usable without asking for permission. That's great when you're in a flow state and need output fast. The risk is that it can be confidently wrong — it invented a stat ("40%") that sounds real but isn't sourced anywhere. If you don't catch that, you're sending fake data to prospects.
Claude is a strategic thinker. It slows down slightly to understand what you're actually trying to achieve. That clarifying question it asked wasn't annoying — it was smart. Claude is trained to be more cautious about assumptions, which means its outputs tend to be more logically sound, even if they take a beat longer to arrive. For high-stakes writing like sales emails, legal summaries, or client proposals, that caution is a feature, not a bug.
Gemini is a surface polisher. It's excellent at making something look better without necessarily making it work better. It's trained on an enormous amount of web content, which means it's very good at matching the format of professional writing. But format isn't persuasion. A cold email that looks clean but says nothing still gets deleted.
The mental model to take from this: think of ChatGPT as your fast first draft machine, Claude as your strategic editor, and Gemini as your formatting and polish layer. Used in sequence, they're more powerful than any one alone.
How to Actually Use This to Rewrite Your Own Sales Emails Today
You don't need to just watch this experiment — you can run the same process on your own emails in under 20 minutes.
Step 1: Pull out your weakest-performing email. This is the one with the lowest open or reply rate. If you don't track that, just pick the most recent cold email you sent that got no response.
Step 2: Go to Claude first. Paste the email and use this prompt: "Rewrite this cold sales email. Identify the single biggest weakness in the original before you rewrite it, then fix that weakness as the centerpiece of your version. Add a subject line. Keep it under 150 words." That instruction — identify the weakness first — forces Claude to show its reasoning, which makes the output dramatically more useful.
Step 3: Take Claude's rewrite to ChatGPT. Use this prompt: "Here's a sales email. Make it sound more direct, confident, and human. Remove any phrases that sound like marketing copy. Don't change the core message." ChatGPT will sharpen the edges without losing the strategic frame Claude built.
Step 4: Do a final read yourself. Ask: Does this email open with their problem or my product? Does the subject line make me curious? Is the call-to-action one clear step? If yes to all three, it's ready to send.
This four-step workflow takes about 15 minutes and will outperform 90% of cold emails being sent today. The difference between a 2% reply rate and a 12% reply rate often comes down to exactly the kind of fixes these AIs can make.
The Part Most People Get Wrong
Most people use AI to polish their emails. That's wrong. Polishing a broken email just gives you a shinier version of something that won't work.
The real job is diagnosis before rewrite. The original email I used wasn't failing because of grammar — it was failing because it was built around the wrong frame. Telling an AI to "make this better" without identifying why it's bad is like asking a mechanic to make your car "run nicer" without mentioning the engine is missing a cylinder.
This is why Claude won the test. It naturally moved toward diagnosis. When you prompt any AI to identify the problem first, you force it into a more useful mode. Try adding "Before you rewrite, tell me what's wrong with this email in one sentence" to any writing prompt — the quality of the output jumps immediately.
The second mistake: treating every AI output as final copy. Every single rewrite from all three AIs needed at least one human edit — either to remove a made-up stat, adjust the tone, or tighten a sentence that went soft. AI gets you to 85%. You close the last 15%.
Key Takeaways
- Claude wins on strategy: It's the best AI for high-stakes rewrites because it diagnoses problems before it fixes them — not just polishes the surface.
- ChatGPT wins on speed: It produces confident, usable output fast, but always fact-check any stats or numbers it invents.
- Gemini underperforms on precision tasks: It excels at formatting and matching professional tone, but it can miss the strategic core of a brief.
- Diagnosis beats polish: Always ask AI to identify the problem before the rewrite — this single habit improves output quality more than any other technique.
- The real email framework: Every strong cold email opens with the reader's problem, not your product — and every AI rewrite should be judged against that standard first.
What to Do Right Now
Open Claude and paste the last cold email you sent that got zero replies. Use this exact prompt: "Before you rewrite this, tell me in one sentence what the biggest problem is. Then rewrite it to fix that problem specifically. Add a subject line. Keep it under 150 words." Do it in the next 10 minutes — you'll have a better email before you finish your coffee, and you'll understand all three AIs more clearly from watching just one of them work.