ChatGPT’s new image generator is really good at faking receipts

This month, ChatGPT rolled out a new image generator with its 4o model, and it's pretty darn good at putting text into images. Folks are already using it to whip up fake restaurant receipts, adding yet another trick to the bag of AI deepfakes that scammers love to use.

Deedy Das, a big name on social media and in the VC world, shared a snap on X of a fake receipt from a real San Francisco steakhouse, claiming he made it with 4o.

You can use 4o to generate fake receipts.

There are too many real world verification flows that rely on “real images” as proof. That era is over. pic.twitter.com/9FORS1PWsb

— Deedy (@deedydas) March 29, 2025

Others jumped on the bandwagon, creating their own fakes, some even adding food or drink stains to make them look more legit:

I think in the original image the letters are too perfect and they don’t bend with the paper. They look like hovering above the paper. Here is my attempt to make it more realistic. Let me know what you think. pic.twitter.com/EixRSHubeY

— Michael Gofman (@michaelgofman) March 29, 2025

The most convincing fake we found at TechCrunch came from France, where a LinkedIn user posted a crumpled-up AI-generated receipt from a local restaurant chain:

We at TechCrunch gave 4o a whirl and managed to create a fake receipt for an Applebee’s in San Francisco:

Still, it wouldn't take much for a scammer to tweak a few numbers using photo editing software or maybe even more specific prompts.

It's pretty clear that making it super easy to generate fake receipts opens up a whole world of fraud possibilities. It's not hard to picture bad actors using this tech to get "reimbursed" for completely made-up expenses.

Taya Christianson, a spokesperson for OpenAI, told TechCrunch that all images made by ChatGPT come with metadata showing they were created by the AI. Christianson also mentioned that OpenAI "takes action" against users who break its usage policies and that they're "always learning" from real-world use and feedback.

When we asked why ChatGPT lets people make fake receipts in the first place, and if that's okay under OpenAI's usage policies (which ban fraud), Christianson said that OpenAI's "goal is to give users as much creative freedom as possible." She pointed out that fake AI receipts could be used for non-fraud purposes like "teaching people about financial literacy" as well as creating original art and product ads.