I Gave All Three AIs the Same Terrible Email Draft — Here's What Happened
Most AI comparisons are useless. They test things like "write a poem about the ocean" and declare a winner based on vibes. I wanted to know something actually useful: which AI can take a weak, rambling email and turn it into something that gets a response?
So I wrote the worst version of a real email — the kind you draft at 11pm when you're tired and just want to send something. Then I gave it to ChatGPT, Claude, and Gemini with identical instructions. The differences weren't small. One AI rewrote it into something I'd actually send. One played it safe. And one made it worse in a way that's going to change how you think about AI writing tools.
Here's exactly what I did, what each AI produced, and what it tells you about picking the right tool for the right job.
The Terrible Email I Started With (And Why Bad Input Matters More Than You Think)
Here's the original draft I gave all three AIs — no editing, no cleanup:
"Hey, I wanted to reach out because I think there might be an opportunity here for us to maybe work together on something. I've been following your work for a while and think it's really great. I know you're probably busy but if you have time it would be cool to chat. Let me me know what you think. Thanks"
This email has every common problem: vague ask, no hook, hedging language ("maybe," "might," "if you have time"), and zero reason for the recipient to say yes.
The prompt I gave each AI was identical: "Rewrite this email to be professional, persuasive, and concise. Keep it under 100 words. The goal is to get a response."
Same input. Same instructions. Completely different outputs.
Before we get to the results, here's why this test matters. Most people use AI as a proofreader — they ask it to "clean up" their writing. That's leaving 80% of its value on the table. The real skill is knowing which AI to use when the job requires tone intelligence, not just grammar fixes.
What ChatGPT, Claude, and Gemini Actually Produced
ChatGPT (GPT-4o) came back fast with this:
"Hi [Name], I've been following your work on [topic] and genuinely admire what you've built. I'd love to explore a potential collaboration — I think there's a real opportunity to [specific outcome]. Would you be open to a 20-minute call this week? I'll keep it focused and make it worth your time. Best, [Your name]"
Clean. Professional. But notice it leaned on placeholders like "[topic]" and "[specific outcome]." ChatGPT preserved the structure of persuasion without filling in the gaps — it rewrote the skeleton but left the muscle for you to add.
Gemini produced the longest response of the three, coming in at 142 words despite being explicitly told under 100. It opened with "I hope this message finds you well" — arguably the most ignored sentence in the history of email. The rewrite was polished in a corporate, newsletter-friendly way, but it lost all personality and still didn't have a clear, single ask.
Claude did something different. Its output:
"Hi [Name], Your work on [topic] caught my attention — specifically [one specific thing]. I have an idea for a collaboration I think could benefit us both. Can we get 20 minutes on the calendar this week? I'll come prepared with specifics. — [Your name]"
That last line — "I'll come prepared with specifics" — is doing serious persuasion work. It removes the recipient's risk. It signals you're not going to waste their time. That one sentence is the difference between a reply and silence.
Claude won this round. Not because it's "smarter," but because it understood that a cold email's job isn't to impress — it's to reduce friction.
The Hidden Difference: Tone Intelligence vs. Template Intelligence
Here's the thing most AI tool comparisons miss entirely: there's a difference between tone intelligence and template intelligence, and these three AIs have very different strengths.
Template intelligence is knowing the structure of a good email — hook, value prop, clear CTA, short paragraphs. Every major AI has this now. It's table stakes. ChatGPT and Gemini both demonstrated solid template intelligence in the test above.
Tone intelligence is knowing why a specific word choice changes how a reader feels, and editing toward that. It's the difference between "Let me know if you're interested" and "Can we get 20 minutes on the calendar?" One invites a non-answer. The other invites a yes or a no — both of which are better than silence.
Claude consistently shows stronger tone intelligence in persuasive writing tasks. This isn't a coincidence. Anthropic built Claude with what they call Constitutional AI, a training approach that emphasizes nuanced reasoning about language and ethics. That shows up in writing as an unusual sensitivity to subtext.
A useful mental model: think of ChatGPT as a brilliant generalist who can write anything competently, Gemini as a research-first writer who's better when you need depth and sourcing, and Claude as your sharp editor who asks "but is this actually going to work on a real human?"
When the job is persuasion — sales emails, pitch decks, negotiation messages — start with Claude. When you need breadth, structure, or integrated data, go with the others. Knowing this saves you 20 minutes of re-prompting every single time.
How to Use This Today: The 3-AI Email Workflow That Gets Results
You don't have to pick one AI and commit. The best results come from using all three in sequence — each one for what it does best. Here's the exact workflow.
Step 1 — Draft in ChatGPT. Paste your rough email and use this prompt: "Rewrite this email with a clear hook, a single specific ask, and an ending that makes it easy for them to say yes. Under 80 words." ChatGPT is fast, reliable, and great at generating a solid first pass with strong structure.
Step 2 — Sharpen in Claude. Take ChatGPT's output and paste it into Claude with this prompt: "Read this email as the recipient. What's the one line that would make you hesitate to respond? Fix it. Then tell me if the ask is clear enough." Claude will flag the weak spot every time. This step alone has made my outreach emails measurably better.
Step 3 — Gut-check in Gemini. Paste the final version into Gemini and ask: "Does this email make any assumptions about the reader that could feel presumptuous or off? Suggest one alternative opening if so." Gemini's broader training data makes it useful for flagging cultural tone issues or assumptions you didn't realize you were making.
The whole process takes under 10 minutes and you end up with an email that's been stress-tested by three different language models. That's the kind of editorial process that used to require a senior copywriter. Now you have it on your laptop for free.
One more thing: save your best outputs as personal templates. After you run 10 emails through this workflow, you'll start to see patterns — your strongest openers, your best CTAs. That's a goldmine of your own voice, refined by AI.
The Part Most People Get Wrong
Most people treat AI like a vending machine. They put in a bad prompt, get a mediocre output, then conclude the tool "doesn't work." That's wrong — and it's costing them results daily.
The real mistake is under-specifying the context. Saying "rewrite this email to be better" is like telling a tailor "make my suit fit." Better how? For who? With what goal? When you give Claude or ChatGPT zero context about your recipient, your relationship with them, or the desired outcome, you'll get generic output. Every time.
The fix is a single extra sentence before your prompt. Compare these two: "Rewrite this email to be better" vs. "Rewrite this email — I'm a freelance designer reaching out to a startup founder I've never met, asking for a 20-minute intro call. They're busy and skeptical of cold outreach." The second version will produce something 10x more usable, and it took you 15 extra seconds to write.
And stop judging an AI's capability by its first response. The best outputs almost always come from one follow-up prompt: "The tone is right but it's still 20 words too long — cut the least important sentence" or "The opening is weak — give me three alternative first lines." That follow-up is where the real quality lives.
Key Takeaways
- Tone intelligence: Claude currently leads on persuasive writing because it reasons about how words land on a real reader, not just whether they're grammatically correct.
- Template intelligence: ChatGPT produces the most reliable structure and is the best starting point for any email rewrite.
- 3-AI workflow: Draft in ChatGPT → sharpen in Claude → gut-check in Gemini gives you better results than any single tool alone.
- Context is everything: Adding one sentence of recipient context to your prompt will improve AI output more than switching tools.
- Follow-up prompts: The best AI output rarely comes on the first try — always send at least one refinement prompt before you use what it gives you.
What to Do Right Now
Open your email drafts right now and find the one you've been putting off sending because it doesn't feel quite right. Paste it into Claude with this prompt: "Rewrite this email to be under 80 words, with a clear single ask and a closing line that makes it easy to say yes. Tell me what was weakest about the original." Send the result within 10 minutes — before you talk yourself out of it.