How to Analyze CSV or Excel Data With ChatGPT and Verify the Results
Analyze CSV and Excel files with ChatGPT: inspect columns, check missing data, reproduce calculations, and make charts without invented numbers.
Content version: 2026-11-07 · Data and spreadsheets
Quick answer
Start by inspecting the dataset's columns and row counts, define an exact question, ask ChatGPT to show the calculation or code behind each result, and independently verify a sample of totals before using charts or conclusions.
Copy a CSV analysis templateUploading a dataset does not make every number trustworthy. Column names can be ambiguous, missing values can change the denominator, and charts can make weak conclusions look precise.
This workflow focuses on analysis you can reproduce. Remove personal identifiers, payment details, and credentials before uploading data, and use an approved environment for sensitive datasets.
Step-by-step workflow
1. Create a data dictionary before plotting
Ask for column names, types, units, unique values for categories, missing-value counts, suspected duplicates, and date coverage. Compare the reported dimensions with the original file.
Example: Inspect these columns only. Report shape, types, null rates, duplicates and ambiguities. Do not draw conclusions yet.2. Translate a question into a precise metric
Define what is being counted, what groups to compare, how missing rows are handled, which date window is used, and what outcome would answer the question.
Example: For each calendar month, show unique orders and total revenue, excluding refunded orders. State the currency and whether returns are represented.3. Require executable or auditable calculations
Request the transformation steps or code, aggregation rules and example arithmetic. When code execution is unavailable, label outputs as illustrative rather than computed.
Example: Show the pandas steps and the grouping formula. If you cannot execute against the actual file, do not report numerical results.4. Check denominators and edge cases
Recalculate at least one subtotal, grand total and percentage outside the model. Check date parsing, currency symbols, categorical spelling variants and outliers.
Example: For this percentage, identify numerator, denominator, excluded rows and rounding rule.5. Choose the chart from the question
Use a line chart for a time trend, bars for category comparisons, scatter plots for two numeric variables and pie charts only for a few parts of one whole. Include units, period and caveats.
Example: Create a monthly revenue line chart with the same filters used for the summary table; include the source row count.
Mistakes to avoid
- Do not let a model invent columns, numbers or correlations it did not compute.
- Do not treat correlation as a causal explanation; check confounders and definitions.
- Do not publish a chart without labels, units, filter criteria and a way to reproduce it.
Common questions
Can ChatGPT calculate values from my file?
Data analysis capabilities vary by model, plan and tool availability. Ask to see the actual processing steps and verify a sample rather than assuming the model executed code.
What chart should I use for monthly data?
A line chart is usually appropriate for a continuous time series. Use bars when comparing discrete months or separate categories matters more than the trend.
What should I do with missing values?
Measure their amount and distribution before choosing exclusion or imputation. Document the rule because it can materially change totals and comparisons.
Continue with related resources
Further reading
Explore the official documentation for more details about the tools and methods discussed.