Fake Receipt Detector
AI image tools can now produce a convincing restaurant bill, shop receipt or invoice from one sentence of instructions. Upload the receipt and FauxLens checks it for AI generation and edits, and reads the watermark OpenAI and Google put in their images.
Upload the image you were sent. It is deleted after the check.Drop the receipt here or click to upload. It is deleted after the check.
JPG · PNG · WEBP · HEIC
Why trust the answer, and where it stops
How it worksWhat it cannot tell you
Forwarded and heavily compressed screenshots hide some of the signs we look for. A clean result is about the image. It does not mean a payment or a person is genuine.
We read the AI watermark
OpenAI and Google hide a watermark in images their AI tools make. When it is there, the report names whose tool made the image.
You see the evidence
Each result lists what was found, with a link you can send to the other person, a marketplace or your bank.
How to check whether a receipt is real
Do the math
Add up the lines and check the subtotal, tax and total. Generated receipts often carry figures that look right and do not add up.
Check the tax rate and currency against the location
Sales tax, VAT or GST should match the place the receipt claims to come from, and the currency and date format should fit the country.
Look up the merchant
Does the address exist, does the phone number match, and was the business open on that date and at that time?
Check the payment line
Card type, the masked card digits and the terminal details should look like what that merchant’s register actually prints.
Look at the paper
On a photographed receipt, creases, shadows and print fade should agree with each other. AI-made receipts often have texture that is too even up close, with text that stays perfectly sharp across a fold.
What a fake looks like
We made this example with an AI image tool for this page. No real person or payment is involved.
AI-GeneratedFauxLens result on the original file
The math gives it away: $81.50 plus $7.34 in tax is $88.84, not $95.84, and Portland charges no sales tax. The scan flagged it from its hidden content credentials.Where fake receipts are used
Expense claims, returns and refund fraud, proof of purchase in a marketplace dispute, warranty and insurance claims, and "proof of payment" for a deposit. Anywhere a receipt stands in for a transaction someone else cannot see, it is worth faking, and since 2025 the cost of faking one has dropped to a single prompt.
Why AI receipts are the easier case
Image models now render legible text well enough to produce a realistic receipt, complete with crumpled paper and thermal-print fade. That is a new problem for expense and returns teams, and it plays to what FauxLens checks. Images from OpenAI's and Google's tools carry an invisible watermark designed to survive screenshots and compression, and when it is detected the scan names which company's tool made the image. Without a watermark, the detection model still scores the image for AI generation.
Read moreShow less
A photo of a real printout that was edited afterwards to change a total is harder. A careful edit can pass, which is why the checks above matter as much as the scan.
The check a fake cannot pass
For expense and refund teams, the strongest control needs no image analysis at all: match the receipt to the card transaction. A receipt with no matching charge on the statement, or a charge for a different amount, settles it however the image looks.
Got a receipt to check?
Related checks
Frequently asked questions
Yes. Current image models can produce realistic receipts from a text prompt. Since May 2026, images from ChatGPT carry content credentials and an invisible watermark, and Google's image tools add their own SynthID watermark. The credentials are usually lost when an image is screenshotted or re-saved, but the watermarks are designed to survive that, and FauxLens reads them. Without one, a detection model still scores the image.
It works on images: a photo or screenshot of an invoice is analyzed like any other image. A PDF is not scanned, so check an invoice PDF against the supplier’s known details and your own records.
It means the image shows no sign of AI generation or AI editing. A real receipt edited carefully by hand, or a genuine receipt for a purchase that was later refunded, can still pass. Check the arithmetic and the merchant too.
For spot checks, yes, and the report shows which signal fired. For volume, the developer API runs the same analysis programmatically.
Updated
Keep exploring
Guides
- How to Tell if a Picture is Fake (5 Methods, Free Tools)Not sure if a photo is real? This guide covers five proven methods to check if a picture is fake, from visual tells to forensic tools, with step-by-step instructions for each.
- The Rise of Synthetic Identity Fraud: A 2026 CrisisScammers are using AI to create entire fake personas. From bypassing KYC checks to the 'Romance Scam' economy, learn how synthetic identities are reshaping fraud.
More tools
Detect while browsing: try the Chrome extension
Right-click any image · No account required