We Gave ChatGPT, Claude, and Gemini 50 Real Work Tasks — Here's Who Won
Photo by Photo by Visual Karsa on Unsplash on Unsplash
AI companies love to publish benchmark scores. Nobody's job involves answering graduate-level physics questions in a timed test, though. So we took a different approach: we collected 50 real tasks from our own workweek and handed them, one by one, to ChatGPT, Claude, and Gemini.
Every task was scored on a simple 0–2 scale: 2 for a usable result, 1 for salvageable with edits, 0 for a miss. No partial credit debates, no vibes-based scoring. Here's how it shook out.
The Task List
We grouped the 50 tasks into five categories of ten:
- Email and communication: client follow-ups, awkward deadline pushes, meeting recaps
- Document summarization: a 25-page market report, a dense insurance policy, meeting transcripts
- Spreadsheet analysis: a messy sales CSV, a budget with duplicate rows, a pivot-table request
- Coding: a Python script to rename files, a regex fix, debugging a broken API call
- Brainstorming and strategy: campaign ideas, product naming, pros/cons lists for a hiring decision
Each chatbot got the same prompts, in the same order, on the same day, using their paid tiers.
Email and Communication: Claude 17, ChatGPT 16, Gemini 12
This category came down to tone. When we asked for a follow-up email to a client who'd gone quiet for three weeks, Claude nailed the balance — polite but with a gentle nudge. ChatGPT's version was fine but slightly stiff. Gemini read like a LinkedIn post.
On the trickier task — declining a meeting invite from your own boss — Claude and ChatGPT both produced sendable drafts. Gemini over-apologized to the point of insincerity.
Category winner: Claude, by a nose.
Document Summarization: ChatGPT 18, Claude 17, Gemini 13
We fed each bot the same 25-page industry report and asked for a one-page summary with three actionable takeaways. ChatGPT was the fastest and most structured. Claude's summary was arguably better written but took noticeably longer to generate.
Gemini struggled with the longest document, dropping a key section entirely — the kind of omission that could embarrass you in a meeting if you didn't catch it.
On the insurance policy test, all three flagged the same exclusion clause, which was honestly reassuring.
Category winner: ChatGPT.
Spreadsheet Analysis: Gemini 16, ChatGPT 15, Claude 14
Here's where the integrations matter. We uploaded a sales CSV with duplicated rows, inconsistent date formats, and a mystery column labeled only "misc."
Gemini, living inside Google's ecosystem, handled the file most naturally and made the best guess at what "misc" contained. ChatGPT's analysis was equally sharp but required more back-and-forth to get the output format we wanted. Claude was accurate but conservative — it kept asking clarifying questions when we wanted it to just make a reasonable assumption and go.
For quick "what does this data say" questions, all three were serviceable. For cleanup work, Gemini's Sheets integration gave it an edge.
Category winner: Gemini.
Coding: ChatGPT 17, Claude 17, Gemini 10
The file-renaming Python script? All three got it right. The regex fix? All three got it right. Then we handed them a real bug from our own codebase — an API call that returned a 422 error under specific conditions.
ChatGPT and Claude both identified the root cause within two responses. ChatGPT's explanation was more thorough; Claude's fix was cleaner. We called it a tie. Gemini suggested uninstalling and reinstalling dependencies, which was not the answer.
If you write code for a living, this category alone should drive your subscription decision.
Category winner: Tie, ChatGPT and Claude.
Brainstorming and Strategy: ChatGPT 15, Claude 15, Gemini 14
The closest category of the test. Asked to name a fictional project management app, ChatGPT went safe, Gemini went weird, and Claude landed somewhere interesting in the middle.
For the hiring decision pros/cons list, Claude produced the most balanced analysis and the only one that mentioned counterarguments to its own recommendation. ChatGPT gave the most actionable structure. Gemini was fine, forgettable, fine.
Category winner: Tie.
Final Tally
| Chatbot | Total Score (out of 100) |
|---|---|
| ChatGPT | 81 |
| Claude | 80 |
| Gemini | 64 |
Yes, one point. After 50 tasks, we genuinely could not crown a definitive champion between ChatGPT and Claude, and we suspect you couldn't either without a specific use case in mind.
So Which One Should You Pay For?
Our honest read:
- Choose ChatGPT if you want the best all-rounder, especially for data work and structured output.
- Choose Claude if your work is writing-heavy or code-heavy and quality matters more than speed.
- Choose Gemini if you live in Gmail, Docs, and Sheets — its integrations close most of the capability gap.
If you're one of the many Americans paying for two of these subscriptions, our data suggests that's actually rational: ChatGPT and Claude have nearly complementary strengths. But paying for all three? The math doesn't work for anyone outside of, well, a review site.
