Excel to Markdown tables for ChatGPT and Claude

Spreadsheets are where a lot of business data lives, and asking an AI assistant to analyse them is one of the most useful things you can do. But attaching a raw .xlsx file is often slow, expensive in tokens, and sometimes misread.

Converting the sheet to a Markdown table first gives the assistant a clean, compact version it can read precisely.

| SKU    | Product          | Rate  | Stock |
|---|---|---|---|
| WL-101 | Laminate 1mm     | 1,200 | 340   |
| WL-102 | Laminate 0.8mm   | 1,050 | 125   |

How to convert

1. Drop your .xlsx, .xls, .ods or .csv file onto the converter.
2. Each sheet becomes its own Markdown table, under a heading with the sheet name.
3. Completely empty rows and columns are removed automatically.
4. Copy the result, or download it as a .md file.

Prepare the sheet first

A few minutes of tidying makes the AI's answers far more reliable:

• Keep one header row. Merged cells and multi-row headers confuse every tool. Use one row of clear column names.
• Delete what isn't needed. Remove helper columns, notes and sheets that aren't relevant to your question. Every cell costs tokens.
• Show values, not formulas. The converter uses the values Excel last displayed, so open and save the file in Excel first if it contains formulas that haven't been calculated.
• Use consistent units and formats. Don't mix "₹1.2L" and "120000" in one column.
• Unhide what you want included. Hidden rows and columns are still converted. Delete them if you don't want them sent.

How big is too big?

Markdown tables are compact, but a sheet with tens of thousands of rows will still be large. Docuflat shows an approximate token count for each file. For very large tables:

• Filter to the rows you actually care about (one month, one region, one customer).
• Send a pivot table or summary instead of raw transactions.
• Ask the AI to write a formula or a script for you rather than processing every row in the chat.

Ask better questions

Once the table is pasted, be specific:

• "Which five SKUs had the largest drop in stock between these two sheets?"
• "Group these invoices by customer and total them. Show your working."
• "Check this table for duplicate invoice numbers."

Asking the assistant to show its working, or to list the rows it used, makes mistakes easy to spot.