ChatGPT Prompts for Data Analysis: Copy and Paste Library

I use ChatGPT for data work most weeks, mainly to get a quick first read of a spreadsheet before I decide what's worth digging into. It's fast and it's patient, but it'll happily give you a confident chart built on the wrong column if you're vague. So the prompts below are written to stop that: each one names the question, the columns and the output you want.

What ChatGPT can and can't do with your data

These are the points from OpenAI's own help pages that change how you should prompt:

What OpenAI says	What it means for your prompt
It reads spreadsheets (.xls, .xlsx, .csv), PDFs and text files such as .json and .txt	Export from your tool as CSV or XLSX before you start
For many tasks it writes and runs Python in a notebook, and you should review the code and assumptions	Ask it to show the steps it took, every time
Its analysis environment can't make web requests or call other services	Upload everything it needs; it can't fetch your live dashboard
It may not read exact values from scanned files or table screenshots	Never analyse a photo of a table when the numbers matter
CSV files and spreadsheets have a size limit of approximately 50 MB	Split a very large export by month or region
Interactive charts cover bar, line, pie and scatter	Ask for one of those if you want to hover and explore

The catch most people miss: OpenAI says a file can upload successfully and still be too large or messy for a complete analysis. If an answer looks thin, ask which rows it read before you trust it.

Prompts to check and clean the file first

Run one of these before any analysis. Most bad answers start with a messy sheet.

Copy this prompt

I've uploaded [file name]. Before analysing anything, list every column, its data type, how many rows are empty in each, and anything that looks inconsistent.

Copy this prompt

Check this sheet for duplicate rows, dates in mixed formats and numbers stored as text. Show me what you found in a table, then fix it in a new copy.

Copy this prompt

My data has one header row and one record per row. Confirm that's what you see. If not, tell me which rows break the pattern.

Copy this prompt

Explain each column in plain English as if I'd never seen this file. Flag any column whose meaning you're guessing.

Copy this prompt

Create a cleaned version of this file with consistent column names in lower case and no blank rows, and tell me exactly what you changed.

Prompts to summarise and spot patterns

For the first read of a new dataset.

Copy this prompt

Give me a short summary of this dataset: the date range, the main categories, the biggest and smallest values in [column], and three things that stand out.

Copy this prompt

Group the rows by [category column] and show the count and average of [number column] for each group, sorted from highest to lowest.

Copy this prompt

Find any outliers in [column]. Explain how you defined an outlier and list the rows, so I can check them myself.

Copy this prompt

Compare [period A] with [period B] for [metric]. Show the change for each [category] and tell me which ones moved the most.

Copy this prompt

What are the top five patterns in this data that a busy manager would want to know? Keep each to one sentence and point to the rows behind it.

Prompts for charts that make sense

Say which chart you want. Otherwise it picks for you.

Copy this prompt

Make a line chart of [metric] by week from [column]. Label both axes, start the y-axis at zero and add a one-line takeaway under it.

Copy this prompt

Create a bar chart comparing [metric] across [categories], sorted from largest to smallest, with the values shown on each bar.

Copy this prompt

Draw a scatter chart of [column A] against [column B]. Tell me whether there's a visible relationship, and remind me that this doesn't prove one causes the other.

Copy this prompt

Turn this table into the clearest single chart for a slide aimed at [audience]. Explain why you chose that chart type.

Prompts to analyse survey and feedback text

For open-text answers, reviews and support tickets.

Copy this prompt

This sheet holds [number] customer comments in the [column] column. Group them into themes, count how many fall into each, and quote two short examples per theme.

Copy this prompt

Sort these survey answers into positive, neutral and negative. Show the counts and list the comments you weren't sure about.

Copy this prompt

Pull out every suggestion or feature request from these responses and rank them by how often they come up.

Copy this prompt

Read these support tickets and list the five most common problems, with the wording customers use to describe each one.

Prompts to check the answer before you use it

The step that saves you from a wrong slide.

Copy this prompt

Show me the exact steps and code you used for that result, in plain language first, then the code.

Copy this prompt

Recalculate [result] a second way and tell me if the two answers match. If they don't, explain why.

Copy this prompt

Which assumptions did you make in this analysis? List them and tell me which one would change the result most if it were wrong.

Copy this prompt

Did you read every row of the file? Tell me how many rows you analysed and whether any sheets or columns were skipped.

Copy this prompt

Write a short note I can send to a colleague explaining what this analysis shows, what it doesn't show, and what data would make it stronger.

How to get better results from ChatGPT data analysis

• Clean the sheet first: descriptive headers in the first row, one record per row and no empty rows splitting the data, which is what OpenAI recommends.
• Name the columns in your prompt rather than describing them loosely.
• Ask for one task per message. A summary, a chart and a forecast in one go usually means three shallow answers.
• Always ask it to show its method, and check a couple of rows by hand.
• Keep personal or confidential details out of the file unless your workspace allows it.

If your data lives in Excel and you'd rather work there, my guide to using Claude inside Excel covers that route. And before you upload anything sensitive, read my plain answer on whether ChatGPT is private. For recurring reports, keeping the files and instructions together in a ChatGPT project saves you re-uploading every time.

My advice on using ChatGPT for data analysis

Treat ChatGPT as a fast junior analyst, not the final word. It's brilliant for the first look, the tidy-up and the chart you'd otherwise spend twenty minutes formatting. But it chooses methods and chart types on its own if you let it, and OpenAI says that choice doesn't always match your intent. Give it a clean file, a precise question and a request to show its working, and it'll save you hours. Skip those three and you'll get a lovely chart that answers the wrong question.

Frequently asked questions

Can ChatGPT analyse Excel files?

Yes. OpenAI says ChatGPT supports .xls, .xlsx and .csv files, plus PDFs and text files, and can summarise them, build tables and create charts.

What is the file size limit for ChatGPT data analysis?

OpenAI says CSV files and spreadsheets have a size limit of approximately 50 MB, depending on the size of each row. A file can upload and still be too complex to analyse fully.

Can ChatGPT read data from a screenshot of a table?

Not reliably. OpenAI says it may not extract exact values from image-based tables or scanned files, so upload a spreadsheet when exact values matter.

Can ChatGPT pull data from the web for analysis?

Not inside its analysis environment. OpenAI says that Python environment can't make web requests or call other services, so upload or connect the data first.

How do I know ChatGPT's analysis is right?

Ask it to show the steps and code it used, recalculate key results another way, and check a few rows yourself before you rely on the answer.