‘The era of fraud is here’: ChatGPT is now great at faking receipts

ChatGPT’s latest image generator, powered by the GPT-4o model, was released on March 25, and has been both lauded and criticized for its advanced design production. Aside from recreating Studio Ghibli content, some users of the Artificial Intelligence (AI) tech realized the system update had other, more nefarious capabilities: creating fake receipts.

“You can use 4o to generate fake receipts,” X user @deedydas wrote in the caption of their post. “There are too many real world verification flows that rely on ‘real images’ as proof. That era is over.”

The post includes an image of what looks like a receipt for a restaurant called the Epic Steakhouse. The receipt includes the location of the restaurant, an itemized list of items, the subtotal, tax and tip, and a bill totaling $277.02.

Posted on March 29., 2025, the fake steakhouse receipt has over 2.3M views, 11K likes, 4.3K saves, 1K reshares, and 811 comments.

In a follow-up post, @deedydas included the exact prompt used to create the image.

“‘Generate me a photorealistic iPhone picture of a $277.02 wrinkled receipt on a wooden table with reasonable numbers. Make the math add up. The restaurant name is X and the address should be Y.’”

Epic Steakhouse is a real restaurant, located in San Francisco. Some foodie followers were able to spot the fake based on the price of the menu items.

“You can tell it’s fake by the fact there is nothing at Epic that’s < $10 :),” wrote X user @amasad in the comments, to which @deedydas replied, “Hahahaha, facts.”

Others, like X user @michaelgofman, tried their hand at refining the image.

“I think in the original image the letters are too perfect and they don’t bend with the paper,” the wrote in the comments of the post. “They look like hovering above the paper. Here is my attempt to make it more realistic. Let me know what you think,” with a new image of the receipt that had some food stains and a grainier look.

What can fake ChatGPT receipts be used for?

Many social media users were curious how an AI-generated receipt could be used convincingly. Several offered takes on how these fake receipts could be employed (spoiler alert: all of them involve fraud).

“People still playing checkers you could use it to make receipts for tax reasons,” wrote Instagram user @airborne11b on a post roundup from Instagram account @NotSoSecretMenu, adding that it’s “probably illegal.”

“Oh I’m thinking about any workplace that allowed you to send documentation by direct messages on WhatsApp,” added Instagram user @zioncityrd.

Responding to @deedydas’ post, X user @heathermhuang pointed out that “Those receipt uploading reward apps may have trouble now,” while X user @AnshulGarg1986 added that “This will become a nightmare for companies reimbursing expenses.”

OpenAI says creating fake receipts could be ‘useful’

Can AI companies be held responsible for fraud that occurs from users generating fake receipts with their technology? The answer isn’t entirely clear, but a spokesperson for OpenAI stated in an article for TechCrunch that, “all of its images include metadata indicating they were made by ChatGPT” and that OpenAI “‘takes action’ when users violate its usage policies,” which explicitly states it bans fraud.

New research claims that AI can actually be used to defraud fraudsters, however doing so requires proper training and expertise. Still, OpenAI’s spokesperson mentioned that their technology is “‘always learning’ from real-world use and feedback” which will, in theory, eventually prevent fraudulent imaging.

Is there any chance this kind of image generation could be useful? OpenAI’s spokesperson flipped the script on the topic, adding that the company’s “‘goal is to give users as much creative freedom as possible’” and that this kind of usage could be turned into teachable moments, especially when it comes to training in fraud-spotting, or using fake AI receipts as visual examples for financial literacy education, art projects, or advertising.

Netizens are fighting back against fake ChatGPT receipts

There are ways that companies can combat receipt fraud, but the bottom line is clear: AI will make it much easier to generate these kinds of receipts, at a much faster rate than Photoshop or other image-alteration methods could.

“The era of fraud is here,” wrote X user @DataChaz on Apr. 2., 2025. “This is a fake receipt generated by GPT-4o. Every detection tool has failed.”

The social media user included a screenshot of an AI-generated receipt that was run through what appears to be an AI-detection system. The system reported that “The input is not likely to contain Al-generated or deepfake content,” detecting only 26.8% of the image as fake.

@DataChaz challenged others to run the image through their own systems, and noted that some people in the comments pointed out that “the numbers don’t quite add up, there’s a typo in the menu, etc.,”

While he admitted that the technology isn’t perfect, @DataChaz also understood that these kinds of bugs could be “fixed in the next GPT4o update” and that “AI will soon be generating perfectly faked receipts.”

As a result, some are already working on building new technology to fight this kind of fraud. X user @will__ye posted an image of a fake receipt entered into the financial operations platform Ramp, insinuating that the system could be used to “help catch AI-generated receipts like these” using metadata analysis “to automatically detect whether or not it’s real.”

@will__ye also noted that companies can add anti-AI or AI-rejection policies to their approval systems, for extra protection.

The Daily Dot reached out to OpenAI via email, and @deedydas via X DM, for comment.

The internet is chaotic—but we’ll break it down for you in one daily email. Sign up for the Daily Dot’s web_crawlr newsletter here to get the best (and worst) of the internet straight into your inbox.