How Do You Use ChatGPT for Excel?

How Do You Use ChatGPT for Excel?

For about three years, "ChatGPT for Excel" meant copying a formula out of a browser tab and pasting it into a cell, then fixing the range references by hand because the model could not see your workbook.

That is no longer the only option. OpenAI announced an official ChatGPT for Excel add-in on 5 March 2026, and by 5 May it had gone generally available for Excel and Google Sheets alike. It sits in the Excel ribbon, reads the actual cells in front of it, and edits the workbook directly.

Which means there are now two distinct ways to use ChatGPT with a spreadsheet, and they suit genuinely different jobs. Picking the wrong one is the main reason people conclude this does not work.

The Short Answer

There are three routes, in descending order of how closely the model is coupled to your file:

1. The ChatGPT for Excel add-in. Install it from Home → Add-ins, sign in, and describe what you want in plain language. It builds and updates live models, runs scenarios, traces errors and explains how assumptions flow through the sheet.
2. Upload the file to chatgpt.com. Attach an .xlsx or .csv and ask questions about it. ChatGPT runs Python over the data, produces charts and tables, and hands you a written analysis. It does not edit your original file.
3. Ask for a formula in the chat window. The old method. Still the fastest way to get a single XLOOKUP or a nested IF when you do not need the model to see anything.

Most people want the add-in for building and the upload route for analysing.

Spending hours in spreadsheets when the data should be telling you something? Get a free SEO audit and we will turn your search data into something you can act on.

What the ChatGPT for Excel Add-In Actually Does

OpenAI describes it as an add-in that brings ChatGPT directly into workbooks. In practice it does four things that the browser version cannot.

It sees your cells

The model reads the workbook, so a request like "add a column that flags any customer whose spend dropped more than 20 per cent quarter on quarter" works without you describing your column layout first. OpenAI's own framing is that it links its answers back to the cells it referenced, so you can check where a number came from.

It edits the file

This is the real shift. It writes formulas into cells, adds columns and restructures ranges. OpenAI says it asks permission before editing, so changes are reviewable and undoable — which you should treat as a feature to actually use rather than a reassurance to skim past.

It traces errors

Hand it a #REF! or a circular reference and it works backwards through the dependency chain. On a model somebody else built, this is the single most valuable thing in the list.

It runs scenarios

Change an assumption and ask what happens downstream. It explains why outputs moved, rather than just moving them.

The add-in launched in beta on GPT-5.4 and moved to GPT-5.5 at general availability. OpenAI also shipped a set of financial data integrations alongside it — Moody's, Dow Jones Factiva, MSCI, Third Bridge and MT Newswires among them — plus support for connecting proprietary data through the Model Context Protocol.

How to Install and Use It, Step by Step

1. Install the add-in

In Excel, go to Home → Add-ins, search for ChatGPT, and install. It appears in the ribbon.

For Google Sheets the path is Extensions → Add-ons → Get add-ons, then search for ChatGPT.

2. Sign in

Open the add-in and sign in with your OpenAI account. On Enterprise, Edu and Teacher workspaces the feature is off by default and an administrator has to enable it, with role-based control over who gets access. If it is missing and you are on a work account, that is the reason.

3. Check it can see the right range

Select the data you want it working on before you ask. Being explicit about the range costs you four seconds and prevents the most common category of wrong answer.

4. Ask for the outcome, not the mechanism

"Build a 12-month cash flow projection from the assumptions in column B" works better than "write me a formula." You are talking to something that can see the sheet; describe the destination.

5. Review every edit before accepting

The add-in asks before writing. Read what it proposes. This is not a formality — a model that can restructure ranges can restructure them wrongly.

6. Spot-check the arithmetic

Pick two or three output cells and verify them by hand or with a quick independent calculation. If you want the longer version of why this matters, we have written about whether ChatGPT can actually do maths and where it reliably goes wrong.

Guessing at your numbers is expensive. Get a free SEO audit and start with data you can trust.

When to Upload a File Instead

The add-in builds. The upload route analyses. For exploratory work on a dataset you are not trying to restructure, uploading is usually better.

Attach a spreadsheet at chatgpt.com and ChatGPT will summarise rows, columns, trends and outliers, build tables for row-by-row review, chart the data, and run Python-based transformations and statistical analysis. OpenAI's documentation is upfront that you should review the generated code and outputs before relying on them.

Supported formats include .xls, .xlsx and .csv, alongside PDFs and text and data files such as .json, .xml, .yaml, .txt and .md. Exact support varies by model, plan and workspace settings.

On limits, OpenAI's current guidance is a hard ceiling of 512MB per file, with spreadsheets and CSVs exempt from the two-million-token text cap but capped at roughly 50MB depending on row size. Upload allowances are per-plan and change often enough that it is worth reading the live limits page rather than trusting a number in an article.

Which route for which job

• Building or fixing a model: the add-in.
• Understanding a dataset you were just sent: upload it.
• One formula, no context needed: just ask in the chat window.
• A recurring monthly process: the add-in, because the file stays where it lives.
• Data you cannot send to a third party: neither, until you have checked your organisation's policy.

The same principle applies to your search data: the tool is only useful if you can tell when it is wrong. Get a free SEO audit and get a read on your site you can actually check.

Limitations Worth Knowing Before You Rely On It

OpenAI's own FAQ lists a set of Excel features the add-in does not support, and the list is more significant than it first appears:

• Office Scripts
• Power Query
• Pivot tables
• Data validation
• Slicers
• VBA macros

Read that again if you work in finance or operations, because it rules out a large share of real-world Excel. If your workbook's logic lives in Power Query or its reporting layer is built on pivot tables, the add-in cannot work on the part that matters. You are back to uploading a file or writing the steps yourself.

Large workbooks may not fit. OpenAI notes that very large files may exceed the context window, producing partial results. The failure mode here is quiet: you get an answer about the portion it could see, with no loud warning that it only saw a portion.

It can edit your workbook. Which is the point, and also the risk. Work on a copy for anything important.

Plan availability is a moving target. OpenAI's own pages have carried inconsistent plan lists through the beta and GA transition. Check the live page rather than assuming.

It does not know your business rules. A formula can be mechanically correct and still wrong, because the model does not know that your fiscal year starts in April or that the "returns" column excludes exchanges.

There is no audit trail by default. A colleague opening the workbook next month cannot see which formulas a model wrote and which a person did. If the file matters, note it yourself in a comment or a changelog tab.

ChatGPT for Excel vs the Alternatives

Versus Microsoft 365 Copilot in Excel. Copilot is built into Excel by Microsoft and is a paid seat add-on to a Microsoft 365 subscription; ChatGPT for Excel is an add-in from OpenAI tied to your ChatGPT plan. Copilot has deeper native hooks into the Office stack and your Microsoft Graph data. ChatGPT's add-in is the stronger modelling assistant and brings the financial data integrations. If you already pay for one, start there rather than buying both to compare.

Versus Excel's own formula suggestions. Deterministic, instant, free, and limited to pattern completion. Not competing in the same category.

Versus a dedicated AI spreadsheet tool. Several products rebuild the spreadsheet around an AI layer. They can be excellent, but migrating a workbook that other people depend on is a large commitment compared with installing an add-in.

Versus learning the formulas. Still the highest-return option for anyone who works in spreadsheets daily. The model is substantially more useful to someone who can tell when it is wrong, which is an argument for understanding the formulas rather than against it.

Versus a BI tool. If you are rebuilding the same spreadsheet every month to produce the same chart, the problem is not that you need AI in Excel. It is that the reporting belongs somewhere else.

Final Thoughts

"ChatGPT for Excel" has stopped being a workaround and become a product, and the honest summary is that it is very good at the thing it does — building, explaining and debugging live models — while being unable to touch Power Query, pivot tables, slicers or macros at all.

So treat the unsupported list as the deciding factor rather than a footnote. If your work sits inside those features, keep uploading files for analysis and writing the structural work yourself. If your work is formulas, models and assumptions, install the add-in and expect it to save you real time.

Either way, review the edits and spot-check the numbers. A model that can see your spreadsheet is still a model that can be confidently wrong inside it.

Ready to stop guessing which pages are worth your time? Get a free SEO audit and see exactly where your site stands in search.