How to Extract Data From a Screenshot or Photo With ChatGPT

Retyping numbers from a screenshot, a receipt photo, or a chart is tedious and error-prone. ChatGPT can read images, so it can pull the data out and hand it back as a table or CSV you paste straight into a spreadsheet. This works with the image upload on free and paid plans.

• A clear image of the data (screenshot, scan, or photo)
• A ChatGPT account that supports image upload
• A target format in mind (table, CSV, or JSON)

Step 1: Upload a legible image

Attach the image with the plus or paperclip icon. Legibility decides accuracy: crop out clutter, make sure text is in focus, and avoid heavy glare on photos. A sharp screenshot beats a tilted phone snapshot every time.

Step 2: Ask for the exact output format

Tell it precisely how you want the data. Naming the format means you can paste the result without reshaping it. CSV drops into a spreadsheet; a markdown table is easier to read in the chat.

Extract the table in this image as CSV.
Columns: item, quantity, unit price, total.
Keep numbers exactly as shown, do not round.
If any cell is unreadable, write UNREADABLE so I can spot it.

Step 3: Get the structured result

The model returns the data in your chosen format. Because you asked it to flag unreadable cells, anything it could not parse is marked rather than silently guessed.

item,quantity,unit price,total
A4 notebook,12,3.50,42.00
Gel pens (pack),8,5.25,42.00
Sticky notes,20,1.10,22.00

Step 4: Spot-check the numbers

Vision models occasionally swap a digit, especially with low-quality images. Verify a couple of values against the picture, and check that any totals add up. Ask does each total equal quantity times unit price, list any rows that do not as a quick sanity test.

Result

A photographed invoice becomes clean CSV in under a minute, with unreadable cells flagged and totals checked. You paste it into your spreadsheet instead of retyping twenty rows by hand.