40 ChatGPT Prompts for Better Excel Charts

Most “ChatGPT for Excel” prompt lists are useless because they ignore the one thing that actually determines whether you get a good answer: ChatGPT can’t see your spreadsheet. It has no live connection to your workbook. So a prompt like “make my data into a nice chart” fails — there’s no data for it to look at. The prompts below work because each one either tells ChatGPT the structure of your data or asks for something that doesn’t need to see it: a formula, a method, a design decision, a troubleshooting path.

That’s also the honest line between this and Copilot. If you want AI that reads your live workbook and inserts charts directly, that’s Copilot’s job, and it’s better at it. ChatGPT’s advantage is that it’s free (or cheap), it explains its reasoning, and it’s excellent at the thinking around charts — which type to use, why a chart is misleading, how to phrase a formula. Use it for judgment and generation; use Copilot for in-grid execution.

Everything here works in ChatGPT, Claude, or Gemini — the model matters less than the prompt. Replace the bracketed parts with your specifics and paste the result into Excel.

Group 1 — Choosing the right chart

The highest-value use of AI here isn’t building a chart. It’s not building the wrong one. If you’re unsure, our chart chooser guide is the human version of everything below.

1. “I have [describe your columns, e.g. 12 months of revenue for 3 product lines]. What chart type best shows [the comparison you care about], and why?” — Expect: a specific recommendation with reasoning, not a list of every option.

2. “I want to show how a total breaks down into parts that change over time. Pie chart or something else?” — Expect: it should talk you out of pies and toward stacked columns or an area chart.

3. “Rank the best three chart types for showing correlation between two variables, from best to worst, with one line on each.” — Expect: scatter first, with honest trade-offs.

4. “My boss wants a pie chart with 14 slices. Give me a better option and a one-sentence way to explain the switch to them.” — Expect: a bar chart plus diplomatic phrasing.

5. “I’m comparing actual vs target for 8 KPIs on one screen. What chart makes this scannable?” — Expect: bullet charts or a small-multiples layout.

6. “What’s the right chart to show the flow of budget from categories to sub-categories?” — Expect: Sankey or a treemap, with the trade-offs.

7. “Talk me out of a chart entirely — when is a formatted table better than a graph for [my data]?” — Expect: a genuine case for tables when precision matters.

8. “I have one number that changed from last quarter. What’s the least silly way to visualise a single metric?” — Expect: it should question whether you need a chart at all.

Group 2 — Building charts and the formulas behind them

ChatGPT can’t click Insert > Chart for you, but it can hand you the formulas, the helper columns, and the exact step sequence.

9. “Give me the step-by-step to build a [waterfall / Pareto / funnel] chart in Excel 365, using the native chart type first and a legacy method as fallback.” — Expect: numbered steps, modern method leading.

10. “Write an Excel formula that creates a dynamic named range for a chart so it grows automatically when I add rows.” — Expect: an OFFSET or a dynamic-array FILTER/INDEX approach.

11. “I have sales in column B and dates in column A. Give me the formula to pull just the last 12 months into a chart-ready range.” — Expect: a spill formula using FILTER.

12. “Generate sample data I can paste into Excel to practise building a [chart type] — 20 rows, realistic values.” — Expect: a clean table you can copy straight in.

13. “What helper column do I need to turn my flat data into the shape a [dumbbell / bullet] chart requires?” — Expect: the exact intermediate calculation.

14. “Write the SEQUENCE and NORM.DIST formulas to build a bell curve for a mean of [X] and standard deviation of [Y].” — Expect: ready-to-paste dynamic-array formulas (see our full bell curve tutorial).

15. “Give me a combo chart plan: which series goes on the primary axis, which on the secondary, and why.” — Expect: a reasoned axis assignment, not just “put one on each.”

16. “I want a chart title that updates automatically to show the latest month. What formula and setup do I need?” — Expect: a linked-cell title formula.

Group 3 — Formatting and design

This is where AI quietly saves the most time, because formatting decisions are judgment calls it’s genuinely good at.

17. “Suggest a professional, colourblind-safe colour palette (hex codes) for a business chart with 4 series.” — Expect: specific hex values, not vague advice.

18. “My chart looks cluttered. List the five things to remove first to make it cleaner.” — Expect: gridlines, borders, redundant legends — the usual suspects, prioritised.

19. “Rewrite this chart title to be a takeaway, not a label: ‘[Sales by Month]’.” — Expect: something like “Sales grew 22% after the March relaunch.”

20. “What’s the ideal number of data labels for a [line chart with 24 points]? Should I label all, some, or none?” — Expect: a “label the endpoints and peaks only” answer.

21. “Give me a rule for when to start a bar chart’s axis at zero and when it’s acceptable not to.” — Expect: zero for bars, always; more latitude for line charts.

22. “Suggest a consistent formatting checklist I can apply to every chart in a report so they look like a set.” — Expect: a reusable checklist you can save.

23. “How do I make one series stand out in a chart where all the others should recede into the background?” — Expect: the grey-out-the-rest technique.

24. “What font sizes and weights should I use for chart title, axis labels, and data labels in a slide-deck chart?” — Expect: a small, sensible hierarchy.

Group 4 — Fixing broken charts

When something’s wrong, describing the symptom to ChatGPT is often faster than searching. It can’t see your file, so be specific about what you observe.

25. “My Excel chart isn’t updating when I add new data. List the likely causes in order of how common they are.” — Expect: manual calculation mode and static ranges near the top.

26. “My chart is showing dates as numbers on the axis. How do I fix the axis type?” — Expect: the text-axis vs date-axis explanation.

27. “Some of my data isn’t plotting on the chart. Walk me through diagnosing why, from most to least likely.” — Expect: hidden rows, filtered data, and range selection.

28. “My data labels are overlapping and unreadable. Give me three fixes ranked by effort.” — Expect: label position, then leader lines, then thinning them out.

29. “The legend on my chart shows ‘Series 1, Series 2’ instead of real names. How do I fix it?” — Expect: the Select Data > Edit series-name fix.

30. “My chart axis maxes out at a weird number and squashes the data. How do I take manual control of the scale?” — Expect: Format Axis min/max/units guidance.

31. “My chart is completely blank even though there’s data. Give me a diagnostic checklist.” — Expect: empty range, hidden source, corrupted chart object.

32. “Copilot made a chart as a Python image and now I can’t edit the colours by clicking. What are my options?” — Expect: re-prompt, edit the code, or rebuild as a native chart (our Python charts deep dive covers exactly this).

Group 5 — Analysing and explaining

Charts exist to make a point. AI is good at helping you find the point and say it clearly.

33. “Here’s a summary of my chart’s numbers: [paste the values]. What’s the single most important takeaway a reader should get?” — Expect: one clear insight, not a recap.

34. “Given this trend [describe it], write two sentences of caption text that explain what’s happening without overclaiming.” — Expect: measured, non-hyped narration.

35. “What might be misleading about a chart that shows [describe it]? Play devil’s advocate.” — Expect: it flags truncated axes, cherry-picked ranges, missing context.

36. “I have monthly data with obvious seasonality. What should my chart account for so the trend isn’t misread?” — Expect: a suggestion to show a rolling average or year-over-year.

37. “Turn this set of chart takeaways into three bullet points for an executive summary: [paste takeaways].” — Expect: tight, decision-focused bullets.

38. “What follow-up question should this chart make my audience ask — and does my data answer it?” — Expect: it anticipates the next question.

39. “Summarise what this chart does NOT tell us, so I can pre-empt pushback in a meeting.” — Expect: honest limitations.

40. “Given these numbers, is a chart even the right format, or would a single sentence do?” — Expect: it’s willing to say “just write the sentence.”

How to get more out of any of these

Two habits separate good results from mediocre ones. First, give structure, not just intent — tell the model your column names, row counts, and what the data represents, because it can’t peek. Second, iterate — the first answer is a draft. “Now make it more concise,” “give me a legacy-Excel version,” “what would break this?” all sharpen it.

And know the ceiling. ChatGPT is a thinking and generation partner, not a hands-in-the-workbook assistant. For that, you want Copilot. The bigger question — whether any of this makes your own chart skills obsolete — we take on directly in Will AI Replace Your Excel Chart Skills?. Short version: it changes the job, it doesn’t erase it.

Frequently asked questions

Can ChatGPT see my Excel file? No. Unless you paste data in or upload a file to a version that supports it, ChatGPT has no view of your workbook. That’s why the prompts here supply structure or ask for formulas and methods rather than “look at my sheet.”

Do these prompts work in Claude and Gemini too? Yes. They’re model-agnostic. The phrasing matters far more than which assistant you use.

Is ChatGPT better than Copilot for Excel charts? Different jobs. ChatGPT is better for judgment, explanation, and generating formulas or sample data cheaply. Copilot is better because it reads your live workbook and inserts charts directly. Use both for what each does well.

Will pasting my data into ChatGPT expose it? Treat anything you paste into a public AI tool as potentially retained, depending on your settings and plan. For sensitive business data, use prompts that ask for methods and formulas rather than pasting the data itself.

Can it write a chart directly into Excel for me? No — it returns text, formulas, or steps you then apply. For direct in-grid chart creation you need Copilot in Excel.

---

Ready to build the chart once you’ve chosen it? Every named chart type has a step-by-step tutorial on this site — start from Which Chart Should You Use?.