Importing Your Data into ChatGPT

Your Data in ChatGPT: CSV, Excel, and Google Sheets

Every analysis starts with the same move: getting your data into the tool. Whether it’s a CSV exported from your CRM, an Excel workbook from finance, or a Google Sheet, you can load these into ChatGPT in seconds. Yet the difference between a productive session and thirty wasted minutes rarely comes down to the analysis itself. It comes down to the import: a bad encoding choice or a stray header row. Here, we walk through each format and the checks that prevent you from building an argument on misread data.

Supported Formats

ChatGPT accepts the main tabular formats, and the choice matters: each carries more or less information about your data structure.

Database extracts often arrive as TSV files because the tab separator avoids conflicts with commas found in addresses or product labels. Conversely, a manually maintained tracking file will almost always be an Excel workbook, complete with sheets and formulas you must account for. ChatGPT also accepts PDFs and text or data files—JSON, XML, YAML, TXT, Markdown. However, scanned tables or images of tables read poorly. When exact values matter, start with a spreadsheet.

Importing a CSV or Excel File

Importing takes four steps:

1. Start a new conversation in ChatGPT
2. Click the + button in the input area, then choose to add photos and files
3. Select your file from your computer
4. Add a context message—for example: “Here are my Q1 2026 sales; can you summarize them?”

The fourth step is the only one requiring thought, and it’s also the one people rush. The context message tells ChatGPT what these rows represent, the time period, and the intended use. Without it, ChatGPT has to guess. ChatGPT then usually displays an interactive preview of the table that you can expand and browse. For calculations, it writes and executes Python code in a Jupyter notebook, typically using the pandas library. It generally detects headers and data types automatically, which you will verify below.

Preparing a CSV That Reads Smoothly

Always save as UTF-8. In any other encoding, the “é” in “Marché” turns into two garbage characters, and your regional groupings become unreadable just as you think you’re done. Next, ensure the first row contains headers. Placing a report title above the table shifts everything down one row, causing ChatGPT to read your field names as data. Merged cells and blank rows in the middle of the file cause similar damage. If your labels contain commas—“Lyon, 3rd arrondissement”—ensure they are enclosed in quotes, as in any clean export, or prefer a semicolon as the separator and tell ChatGPT to expect it. Otherwise, every comma creates a ghost column. Finally, give your columns clear names, keep one row per record, and include only one table per sheet. These are the three recommendations from OpenAI’s help documentation.

What Excel Brings and What It Leaves Behind

An Excel workbook arrives with its own quirks. If your file has multiple sheets, specify which one to analyze: “Analyze only the Sales_2026 sheet.” Formulas raise a separate issue: analysis generally focuses on calculated values, and a workbook loaded with complex formulas may not be analyzed in full. When calculation logic matters for interpretation, describe it in your message, or work directly in Excel using ChatGPT for Excel, covered below. Do not rely on charts and images in the workbook either: ChatGPT works on cell data and reads values embedded in images poorly. Finally, respect the approximate 50 MB limit for spreadsheets, which varies with row size.

Google Sheets and Excel: Two Other Paths

For a Google Sheets file, the first path goes through the Google Drive app:

1. Connect the Google Drive app via plugins or apps, depending on your ChatGPT version, and authorize access to your Google account
2. Attach the file from Google Drive, or ask ChatGPT to find it
3. ChatGPT reads the data using your account permissions—no export needed

This path shines with live files. Consider a pipeline table updated every morning by three sales reps: when someone adds rows, you rerun the analysis on the current version without exporting again. App availability depends on your plan, your country, and your Google Workspace organization’s rules.

The second path does the reverse: it brings ChatGPT into the spreadsheet. ChatGPT for Excel and ChatGPT for Google Sheets add a sidebar to Excel and Google Sheets across all plans, with limited usage on Free and Go. In Excel, install it from the Add-ins button on the Home tab; in Google Sheets, find it in the Google Workspace Marketplace, then access it via the Extensions menu. The assistant works directly on sheets, formulas, and formatting, making it the right tool for understanding or correcting a workbook. Its conversations remain separate from your ChatGPT history, and it’s best to duplicate the file before making major changes.

Verifying the Import Worked

Never ask your first real question before checking what ChatGPT actually read:

Show the first 5 rows and the total number of rows and columns.

ChatGPT runs a Python script to display a preview of the DataFrame. Compare the number of columns with your original file, then reread the column names. If they look like values (“12/03/2026” as a header), you have a one-row offset. Check type consistency, because a column of amounts read as text will break every average calculation. Finally, spot stray rows—typically a totals row that, if counted as another order, would double your revenue without anything looking wrong.

Handling Large Files

A file may be accepted without being fully analyzed: it might be too large, too complex, or loaded with images or formulas. If a result seems incomplete, ask ChatGPT to examine a specific sheet, set of rows, or columns, or split the file—by quarter or subsidiary—and process the parts sequentially. Filter out unnecessary columns before importing, too, since a CRM export often carries forty fields when you only need six. If you first need a high-level view, a sample is enough to frame the analysis. Finally, keep import limits in mind: 80 files every three hours, and 3 files per day on the Free plan.

Before You Import: Personal Data

A CRM export contains names, emails, and sometimes phone numbers. Before importing, remove columns the analysis doesn’t need: revenue by region requires no customer names. Also check your plan’s terms: on individual plans like Free, Plus, or Pro, your conversations may be used to improve models unless you disable this in data controls; on Business, Enterprise, and Edu, this is not the case by default. Note also that deleting a conversation does not delete a file saved separately in your library.

Your Turn

Your first import.

1. Choose a real export—sales, CRM, tracking—and remove unnecessary personal columns
2. Verify UTF-8 encoding and ensure headers are in the first row
3. Import the file with a context message: what the rows represent, the period, and the use case
4. Request the first five rows and the total row and column counts, then compare with the original
5. Correct the file for any discrepancies before analyzing

Official sources: Uploading files and audio to ChatGPT · Data analysis with ChatGPT · ChatGPT for Excel and Google Sheets · Google Drive app and setup in ChatGPT · How your data is used to improve model performance (OpenAI Help Center)

Key Takeaways

• ChatGPT accepts CSV, TSV, Excel, PDF, and text files; Google Sheets connects via the Google Drive app
• Always verify the import with a preview of the first few rows
• UTF-8 encoding and headers in the first row are essential
• ChatGPT for Excel and Google Sheets works directly in the workbook, including formulas
• Remove unnecessary personal data before importing
• For large files, target, split, or sample the data