'A cautionary tale': How clickbait AI images like 'Shrimp Jesus' became a warning against scammers

What might have started off as a harmless joke is now a warning sign of captivating, clickbait AI-generated images taking over social media and gaming algorithms.
Shrimp Jesus
Posted at 5:55 PM, May 20, 2024

When artist Max Arrington first drew Shrimp Jesus, he never dreamed an AI image of Jesus Christ with shrimp-like arms and legs would become a viral internet phenomenon.

"It's a tweet that I drew, and it's become a phenomenon really," Arrington, an artist from Scotland Neck, North Carolina, told Scripps News. "It was supposed to be like a joke drawing. But then I put it together and it kind of just blew up."

While Arrington started selling his digital art of Shrimp Jesus online two years ago, today you can see countless reiterations on Facebook that were generated by artificial intelligence.

What might have started off as a harmless joke is now a warning sign of captivating, clickbait AI-generated images taking over social media and gaming algorithms.

According to a recent report from the Stanford Internet Observatory, which studied 120 Facebook pages, these AI-generated images are used by spammers and scammers to lure in unsuspecting Facebook users. The report, which has not yet been peer-reviewed, defined spammers as "accounts that were pushing their audiences out to a content farm," and scammers as accounts that "were attempting either to sell products that do not appear to exist, had stolen the pages they operated, or were attempting to manipulate their audiences within the comments."

"Sometimes there are spammers and scammers who recognize that they can use this tool to make images — very heart-wrenching images — very easily, very cheaply," Renee DiResta, a technical research manager at the Stanford Internet Observatory who co-authored the report, told Scripps News.

"And they can use that to really draw people in, in hopes of either scamming them, such as by selling them a product that doesn't exist, or sending them to a content farm or some sort of manipulative website that then they have a kind of a negative experience with," said DiResta.

Scripps News reviewed thousands of Facebook posts with AI-generated images and found that many of them have high levels of engagement. And they're not just images of Shrimp Jesus.

"What you're seeing with a lot of this are images that could be real," DiResta said. For example, one image featured "an elderly woman with a birthday cake, and the post says something like, 'I baked it myself.'" Other posts repeatedly included phrases like "Nobody ever blessed me," or "Made it with my own hands," and encouraged Facebook users to "grade" their work.

"You'll see things that are trying to appeal to sympathy, or making people want to engage with the content," DiResta said.

Experts and researchers have called for better transparency from social media companies as AI-generated images increasingly show up on platforms.

According to the Stanford report, Facebook users are increasingly seeing content they do not follow or search for, also known as "unconnected posts." These posts have tripled from 2021 to 2023, growing from 8% of what users see in their feeds to 24%, according to the report, which cited Meta's "Widely Viewed Content Report."

DiResta says this is likely an outcome of Facebook competing with TikTok.

"If you open TikTok, you don't see posts only from people that you follow. It's constantly pushing you videos that it just thinks you're gonna like. And so other platforms are starting to do this also," she said.

Due to Facebook's algorithm, the more people engage with pages featuring AI-generated images on Facebook, the more likely their friends or the people they're connected to will see this content, too, DiResta said.

OpenAI's ChatGPT app is displayed on an iPhone

Artificial Intelligence

Growing concern that artificial intelligence could pose new cybersecurity threats

Clayton Sandell
9:26 PM, Apr 26, 2024

"And that's because the platform can intuit that there's some sort of either connection on the back end, or some sort of relationship between the pages," she said. "It's very common, even in the realm of completely normal activity, to go follow a page on cooking, and then you start to see gardening, and then you start to see jogging."

If a user, for example, engages with religious posts on Facebook, the algorithm could eventually recommend an AI-generated image of Shrimp Jesus. According to the Stanford Internet Observatory report, citing Meta's Transparency Center, a Facebook post with unlabeled AI images landed in the top 10 most-viewed pieces of content last year.

Meta, Facebook's parent company, announced earlier this year in February the company was working with industry partners on "common technical standards for identifying AI content, including video and audio."

"Scammers use every avenue available to them to defraud people and constantly adapt to evade enforcement," a Meta spokesperson told Scripps News. "Content that purposefully intends to deceive or exploit others for money violates our policies, and we remove this content when it's found."

Within a week after Scripps News questioned Meta about the Stanford Internet Observatory report, many of the highlighted posts were gone.

"We welcome more research into AI and cross-platform inauthentic behavior since deceptive efforts rarely target only one platform. We've reviewed the pages in this report and have taken action against those engaged in inauthentic behavior, and demoted the clickbait sites under our Content Distribution Guidelines," a Meta spokesperson said.

Meta said in Q4 2023 it removed 964 million pieces of spam content and 691 million fake accounts from Facebook. The tech giant also said it planned to add transparency labels to AI content users post on Facebook, Instagram and Threads in May.

Consumers, whether they're buying a product or using a social media platform, want transparency when it comes to AI images. According to a recent report from Getty Images, 90% of consumers globally want to know if an image was created using artificial intelligence.

These labels will be increasingly imperative. Not all AI images on Facebook or other social media platforms including Linkedin look like Shrimp Jesus; many of them appear hyper-realistic, save for some clues like extra hands or fingers.

"Some of the early ones were kids with displaying what was supposedly their art. And it would say, 'It's my child's first competition, what do you think,' right? And of course, the point there is to is to generate engagement, to get attention, and then to make the content hit more people," DiResta said.

"These kids don't exist. This art doesn't exist. But people were engaging," she added. "And then sometimes, some of the unscrupulous page owners were sending messages, direct messages to the people who were kind of falling for it."

As these AI-generated images, video and audio advance, and their creators perfect their craft, anyone is susceptible to their intended purpose — deceptive or not.

"Sometimes people use it for creation, but there's always some kind of dark side," DiResta said.

Not all AI images or art may have malicious intentions, but they will be harder to separate from digital art or even photos of real people if they're unlabeled.

"I think it's hard for a lot of people to see things and be able to make the distinction between digital art from a real artist and AI art," Sarah Frazier, a Florida-based artist who purchased a Shrimp Jesus artwork from Arrington, told Scripps News.

"For me it's listening to music made by computer versus music made by a musician," she added. "It's just different, one has real soul to it."

"I feel like it being used to scam people the way it is, is probably one of the biggest things we should take as a cautionary tale with AI art in all of this," Arrington said.

"But even as great as AI can create an image," he added. "I do think people will always want what they can tell their own kind made, when you can see that something has been made by hand."