OCR on receipts and invoices: how to get the numbers right

Why receipts trip up text recognition

OCR (optical character recognition) is the step that turns a picture of text into text you can search, copy or put in a spreadsheet. It works well on a clean printed letter. Receipts and invoices are harder:

• Small print and narrow fonts on till receipts.

• Faded ink. Many shop receipts are printed on heat-sensitive paper that fades over time.

• Creases and curls that bend lines of text.

• Columns of numbers where one misread digit changes a total.

• Logos, stamps and handwritten notes that sit on top of the text.

The good news: most errors start at the capture stage, and that is the part you control.

Capture tips that make the biggest difference

Google's guidance for its own text recognition is a useful checklist for any tool. It recommends text at least 10 pixels high, pages the right way up, common fonts, and "sharp images with even lighting and clear contrasts." In practice:

1. Flatten the receipt. Press it under a book for a minute, or hold the corners down.

2. Use a dark, plain background for light paper so the edges are easy to detect.

3. Light it evenly. Avoid your own shadow and the glare from an overhead lamp. Daylight near a window works well.

4. Shoot straight on. Hold the phone parallel to the paper. Angled shots squeeze the text at one end.

5. Fill the frame, but keep all four edges in view so the scanner can crop.

6. Long receipts: capture them in overlapping sections rather than one tiny image.

7. Scan soon. The sooner you capture a heat-printed receipt, the more legible it will be.

Use a scanner app rather than the plain camera

A scanner app crops, straightens and boosts contrast for you.

• Google Drive (Android and iOS) has a scan feature. Google describes it as scanning "receipts, letters, and billing statements to save them as searchable PDFs." Google notes the web version of Drive doesn't scan.

• iPhone and iPad: in Notes, tap the attachment button, then Scan Documents, adjust the corners and tap Keep Scan. The Files app has the same option under More → Scan Documents.

• OneDrive (iOS and Android): tap +, then Scan photo. Microsoft points former Microsoft Lens users here, since Lens scanning ended on March 9, 2026. Microsoft notes OneDrive saves scans to the cloud, not locally on your device.

Check the numbers after recognition

Treat every OCR result as a draft. Five checks catch nearly all the errors that matter.

1. Re-add the totals

Add the line items yourself and compare with the subtotal, then add the tax and compare with the total. If they don't match, a line was missed or misread.

2. Watch for look-alike characters

Common swaps on receipts:

• 0 and O, 1 and l or I, 5 and S, 8 and B, 6 and G

• a decimal point lost, so 12.50 becomes 1250

• a minus sign or "CR" (credit) dropped from a refund line

• a currency symbol read as a digit

3. Confirm the date format

The US usually writes dates month first. Canada uses more than one format, and Australia writes day first. 04/05/2026 could be April 5 or 4 May. Check the original before a spreadsheet converts it the wrong way.

4. Check the identifiers

Invoice numbers, purchase order numbers, tax registration numbers and account numbers are long strings with no built-in check a person would notice. Compare them character by character. A single wrong digit can send a payment to the wrong account or make an invoice impossible to match.

5. Read the tax lines separately

Sales tax, GST or HST lines are small and often sit next to similar-looking lines. Make sure the rate and amount were captured as separate values and not merged with the line above.

Organize as you go

• Name files consistently, for example 2026-10-03 Northgate Hardware 84.20.pdf. A date-first name sorts itself.

• Keep one folder per month or per project.

• Back up your scans. One copy on one phone is easy to lose.

Should you keep the paper original?

That depends on why you are keeping the receipt, and the rules differ by country. Some examples from official sources:

• Australia: the ATO says you can keep records on paper or electronically, "including photos of your written evidence," and copies must be "a true and clear copy of the original record." It sets a general rule of keeping written evidence for 5 years from the date you lodge your tax return, with exceptions.

• Canada: the CRA says scanned images can replace paper records if they are made to the national standard it refers to. If a business can't meet that standard, it must keep the paper originals.

• United States: the IRS says to keep records until the period of limitations for the return runs out, which is generally 3 years, with longer periods in some cases.

For receipts tied to taxes, warranties, insurance claims or expense reports, check the rule that applies to you, or ask a qualified professional, before you throw paper away. Your employer or insurer may have its own requirements.

A sample check

A synthetic receipt from "Northgate Hardware" scans as: Subtotal 7S.40, GST 7.54, Total 84.20. You spot the S, correct it to 75.40, and re-add: 75.40 plus 7.54 is 82.94, not 84.20. Back on the original, the third line reads 4.26, not 3.00. With that fixed, the totals match.

When a whole invoice needs to be editable

If you need more than the numbers, for example an invoice you have to revise and resend, see our guide on turning a photo of a document into Word text.

A second pair of eyes on scanned paperwork

EasyToDecode is launching soon with a converter that turns scans and photos into editable Word files, keeps tables as real tables, and flags pages that don't closely match the original so you know what to double-check. It will accept PDFs and JPEG, PNG or WebP images up to 10 MB. See how it works or join the waitlist to hear when it opens.