Garbage AI posts like Shrimp Jesus are destroying Facebook

• AI-generated images are overloading Facebook, and it’s sad to see real people engaging.
• Last week, researchers wrote about the pages that post this garbage. They found rampant bad behavior.
• Forgive us, Shrimp Jesus. 

My pastor never told me about the kinds of Jesus I see on Facebook these days: an adult made out of shrimp; a baby surrounded by three-armed fairies and candles (terrible parenting, Joseph); a crucifix worshipped by kneeling flight attendants around a pool.

These images of Jesus are the kind of artificial intelligence-generated images that are overwhelming Facebook, making the social media site even more depressing.

I’m one of the few millennials I know who regularly uses Facebook, mostly to make friends on expat groups and to furnish a new apartment via Marketplace — the only good part of the site. While I enjoy laughing at the likes of shrimp Jesus, I’m alarmed at the engagement I see on these posts and what it portends for the future of social media.

Years ago, Facebook laid the foundation for today’s AI garbage by de-prioritizing content from legitimate sources like news organizations in favor of posts from family and friends. “We feel a responsibility to make sure our services aren’t just fun to use, but also good for people’s well-being,” CEO Mark Zuckerberg wrote when he announced the change in 2018.

But besides a few older relatives and college acquaintances, nobody I know posts much. Instead, my feed is a mix of group posts, ads, and weird AI images — in my case, mostly of fake houses and resorts, perhaps because I belong to a few travel groups.

The comments section of those images is where things turn dark.

A mix of scammers eager to sell you cryptocurrencies and actual humans comment on these photos. The former I expect in every online forum, but the latter is a sad reflection of low media literacy — and researchers are starting to notice.

Last week, a duo from Stanford University and Georgetown University published a paper — not yet peer-reviewed — about AI images on Facebook, after studying 120 Facebook pages through early March. They wrote that AI-generated images appear because Facebook’s algorithm thinks they’ll lead to more engagement — and some of the posts they tracked did go viral.

Unsurprisingly, the researchers found that some people engaging with this content didn’t seem to realize it was AI-generated, and that the groups behind these pages weren’t up to much good. The researchers wrote that some of the page operators engaged in “unambiguously manipulative behaviors” like stealing pages and adding fake followers to boost their status.

The “Thank Jesus For Everything” page, from which I sourced the visuals in this story, looks exactly like what the researchers found. The page, originally dedicated to showcasing local musicians, was hacked in January 2023, its original owner told me.

After the takeover, the page started reposting the kind of Boomer religious images — think prayers in Comic Sans font over bad clip art — that have long circulated on Facebook. Those images came from another page called VFit Athlete, an Indiana-based gym whose owner told me it, too, was hacked, and which has also since pivoted to AI Jesus. Together, the pages have 130,000 followers — a drop in the bucket of Meta’s 3 billion monthly active users.

The true purpose of these pages remains murky: Beyond the obvious comment scams, they could be boosting page views to use for other purposes later, like boosting misinformation — a cycle we’ve seen on Facebook with other pages that predate generative AI.

Meta did not respond to my requests for comment, including about the hacked pages.

Facebook knows generative AI is a problem. It’s been covered by other media outlets well before last week’s study; 404 Media wrote about it in December. Meta’s president of global affairs said in February that the company is working on generative AI labels it will roll out later this year. Notably, this comes in an election year that’s already been marred by generative AI issues, online and off.

But I don’t want to see labeled generative AI content in my feed — I don’t want any generative AI pictures at all. For an alternate universe, look to Meta-owned Instagram: There, I see pictures and videos from my real loved ones, athletes, and brands I like, interspersed with ads that are, at times, well-targeted enough that I bite.

I haven’t — yet — spotted Cruise Ship Jesus on Instagram, a true blessing.

Do you see weird things on your social media feeds? Email me at mmorris at insider.com.