The idea was simple: I broke my typical workday into 5 missions and asked each of the three the exact same thing. No favouritism. No mercy.
- Write the intro for a blog post (like this one, yes)
- Find SEO keyword ideas for a WordPress client
- Debug a chunk of PHP that was doing absolutely nothing right
- Write a professional email without sounding like a robot
- Summarise a 40-page document nobody wants to read
Mission 1 — Writing a blog intro
ChatGPT: Clean, efficient, a little bland. The intro was fine but read like a LinkedIn post written by someone trying to please everyone. Too polished.
Gemini: Interesting structure, but it kept wanting to explain what it was about to do before doing it. Stop. Write, don’t narrate.
Claude: The most natural. It picked up the casual tone I was going for without me having to repeat myself three times. Point to Claude.
Mission 2 — SEO: finding keywords
Here I wanted something concrete. Long-tail keywords, estimated volume, business relevance. Results:
ChatGPT: Solid list, well segmented. It knows its classics.
Gemini: Connected to the web and pulled recent trends. A real advantage. But the presentation was a mess.
Claude: Good list, well structured, but no real-time data. It told me upfront. I appreciate the honesty.
Mission 3 — Debugging PHP
Broken code is the real test. No fluff, just: “find what’s wrong.”
ChatGPT: Fast, precise, found the bug and explained why it was breaking. Very good.
Gemini: Found it too, but with a 15-paragraph explanation. I wanted the fix, not a lecture.
Claude: Efficient and concise. It even flagged a second potential bug I hadn’t noticed. Nice.
Mission 4 — The professional email
The classic. “Write a follow-up email to a client without coming across as desperate.”
ChatGPT: Correct email, ready to send. A bit generic, but usable.
Gemini: Offered three different versions without being asked. Handy or exhausting depending on your mood.
Claude: Asked a clarifying question before writing. Takes 10 extra seconds, but the result was genuinely tailored to my situation.
Mission 5 — Summarising 40 pages nobody wants to read
A dense client brief, in PDF. Who wins?
ChatGPT: Clear, well-structured summary. It loves lists. That’s obvious.
Gemini: Handled the PDF directly without conversion. Big workflow advantage.
Claude: The most “read” summary — it grasped the stakes, not just the chapter titles. But I had to upload the PDF manually.
So, who annoyed me the most?
Gemini. And I say that without any grudge.
Its big advantage — real-time web access — is real, and for SEO especially that’s not nothing. But its main flaw is the verbosity. It talks too much, over-explains, justifies everything. When I just want something done, it’s exhausting.
ChatGPT remains the safe bet. Fast, reliable, versatile. The equivalent of a trusty hammer in the toolbox.
Claude surprised me with its ability to understand context and ask the right questions. Less flashy at first glance, but often the most useful in practice.
Conclusion (the real one)
No absolute winner. But one clear takeaway: use the AI that fits the task, not the one with the best reputation on Twitter. And above all, test them yourself, on your real work. That’s where the differences actually show.