Hey everyone! I was working on a product announcement email for our latest feature and thought, why not make this a little experiment? 😄 I took the same prompt and ran it through five different AI assistants to see how they'd each tackle it.
I tested:
- Le Chat (Mistral)
- ChatGPT (GPT-4)
- Claude (Anthropic)
- Gemini (Google)
- Copilot (Microsoft)
My prompt was pretty specific: "Write a concise, exciting email announcement for our project management tool's new 'Automated Sprint Reports' feature. Target audience is current users. Tone should be helpful and energetic, not salesy. Include a subject line and a clear call-to-action."
Here’s what I found most interesting:
* **Structure & Creativity:** Claude gave me the most ready-to-send, beautifully structured email right out of the gate. Le Chat was a close second—very clean and professional.
* **Tone & Energy:** ChatGPT and Gemini leaned a bit more promotional than I wanted. Le Chat and Claude nailed the "helpful teammate" vibe perfectly.
* **Conciseness:** Copilot was the shortest, but it felt a bit too bare-bones. Mistral's Le Chat struck the best balance for me—complete but not wordy.
The real standout for Le Chat was its understanding of my "not salesy" instruction. The email it generated focused on saving the user time and providing insights, which is exactly what our feature does.
Has anyone else done a similar side-by-side test for content creation? I'd love to compare notes, especially on how you craft your initial prompts to get the best results.
I threw all the outputs into a quick comparison spreadsheet if anyone's interested—just DM me!
Keep building!
Interesting experiment, but I'm always wary of the polish. A beautifully structured email from Claude might *look* ready-to-send, but that's the trap.
These tools often generate generic, interchangeable corporate-voice copy. The "helpful teammate" vibe you liked probably just means it used a few first-person-plural "we're excited to help you" lines. In my tests, swapping the product name between Claude and Le Chat outputs yields near-identical templates.
The real test is feeding that output into a second prompt: "Make this sound like it came from an actual human at our specific company." That's where most of them stumble.
Good point on the generic output. The corporate tone is a default setting for these tools because it's low-risk.
What I'd be curious about is if you tried the same prompt with different system instructions first. Most of these models have a "system" or "context" field. You can prime them with "You are the lead product manager at a startup. You write in a direct, friendly, slightly technical tone." That usually gets me closer to a usable first draft without needing a second polishing prompt.
The structure is the easy part for them. It's the company-specific voice that's missing, like you said.
Build once, deploy everywhere