Tool Comparison: Same Task in Three AIs
You compare Claude, ChatGPT and Gemini on the same task.
Why this lesson matters
You now know three AI tools: Claude, ChatGPT and Gemini. Each with its own strengths, each with limits. Which one you pick depends on the task, and you only notice that once you put them side by side on work you do yourself.
This lesson gives you three test assignments and the points on which you compare. You write down the outcome yourself. A ranking here would be out of date within a quarter, because all three providers release a new model every few months.
Task 1: Excel analysis
The assignment: analyse sales data and generate insights.
Prompt for all three:
"I have sales data with columns: date, product, category,
quantity, price, region. 5,000 rows. Analyse the data and give:
1. Top 5 products by revenue
2. Revenue per region with percentage split
3. Monthly trend with seasonal patterns
4. A chart of monthly revenue per category"What to look at
| Point | The question you ask yourself |
|---|---|
| Calculation | Does it actually calculate, or does the tool only describe how you would calculate it? |
| Charts | Do you get an image you can download, or a description of a chart? |
| Depth | Does the tool only name what's in the data, or also what stands out? |
| Export | Can you take the result as a file into your own environment? |
| Checking | Can you see how a number came about? |
One difference has nothing to do with taste. ChatGPT with Code Interpreter runs real Python and returns a file you can download. If your data already sits in Google Sheets, Gemini saves you copying back and forth. Those are product properties, and you can look them up.
Task 2: Writing a professional email
The assignment: write a sensitive customer email.
Keep reading
Leave your name and email address and you can read the rest
You get the whole lesson right away, and every other lesson stays open after that. No password, no confirmation email.
Prompt for all three:
"Write an email to a major customer (annual revenue €200K)
who is complaining about repeated delivery problems. They are
threatening to switch to a competitor. Tone: empathetic but professional.
Offer a concrete solution. Max 250 words. Style: Dutch
business, not too formal."What to look at
| Point | The question you ask yourself |
|---|---|
| Tone | Does it read like a Dutch business email, or like a translation from English? |
| Empathy | Is there something that truly reassures this customer, or only a stock phrase? |
| Commitment | Is anything promised that your organisation can actually deliver? |
| Length | Does the tool stick to the 250 words you asked for? |
| Anglicisms | Count the English constructions that don't belong in a Dutch email |
This task is the sharpest of the three, because an email to an angry customer goes out as it is. Read the three versions side by side and ask yourself for each one whether you would send it like that. That judgement is more useful than a score.
Task 3: Generating code
The assignment: build a function with error handling.
Prompt for all three:
"Write a Python function that reads a CSV file,
filters rows on a date range, handles missing values,
and exports the result as JSON. Add error handling
for: file not found, invalid date format, empty dataset.
Add a docstring and type hints."What to look at
| Point | The question you ask yourself |
|---|---|
| Does it work | Paste the code, run it, and see whether a result comes out |
| Error handling | Are the three requested cases really caught, or only mentioned? |
| Documentation | Are the docstring and the type hints there, as you asked? |
| Explanation | Does the tool explain why it approaches something that way? |
| Running it | Can the tool run the code itself, or do you have to? |
What differs structurally
The quality of an answer shifts with every new model. What the tools can do shifts more slowly, and that's something you can base your choice on.
| What you need | Where you find it |
|---|---|
| Run Python and get an image back | ChatGPT with Code Interpreter |
| Work in the files on your own computer | Claude Code |
| Right next to your files in Google Workspace | Gemini |
| Research with references to your own sources | NotebookLM from Google |
| Your own assistant with fixed instructions | Custom GPT in ChatGPT, Projects in Claude |
Conclusion: the right tool per task
The only "best" AI tool is the right tool for the right task.
| Pick this tool | When you... |
|---|---|
| Claude | Write text with tone and nuance, analyse documents |
| Claude Code | Want to work directly in your codebase |
| ChatGPT | Analyse data with charts, generate images |
| ChatGPT | Automate repetitive tasks (Custom GPT) |
| Gemini | Work in Google Workspace, analyse video/audio |
| Gemini | Do source-based research (NotebookLM) |
Every comparison of AI tools is out of date within a few months. That's why this lesson gives you the comparison points and leaves the verdict to you. Put your own outcome in a document with a date and review it every quarter.
Practice exercise
Do your own comparison test. This is the core of the lesson, because only your own outcome is useful.
- 1Pick a task you really have to do this week
- 2Write one prompt and give it unchanged to all three tools
- 3Go through the points in the table above and write one sentence per tool
- 4Decide which tool you'll pick from now on for this type of task, and add the date
- 5Repeat it in three months. The outcome changes, and then you'll know right away
You have 5,000 sales rows and you want a chart you can paste into a presentation. Which tool is the obvious choice?
What is the main conclusion of this tool comparison?
Want to keep your progress, get the practice files and sign up for the free evening? Create a free account. All you do is click a link in your email.
Create a free account