Kelsey Smith takes time when she makes her food content on social media. Smith’s videos appear as if they were professionally produced—the yellowish undertones in the lighting highlight the food, and the well-timed cuts in her how-to videos based on original recipes make for seamless transitions. Her clips often don’t run more than 30 seconds, simply showing off the finished dish and linking to the recipe, which is shared on her website. The videos are almost too perfect—so much so that viewers think that her posts aren’t real, but rather a landing page for the ever-growing AI-generated slop content that is seen on social media.
“I’m constantly asked if I’m a real creator, or if I’m AI,” Smith said. “It’s just funny because I work so hard to make my content high-quality, so it’s not replicable by AI, but now the stuff that you can generate is so well, people can’t tell the difference anymore.”
Over the past two years, the AI media has made its way into the food content space—first in the form of plagiarizing blogs and articles, and now it can create videos that are indistinguishable from those made by real people. The generated content has become so abundant that it has diluted and drowned out the creations of food bloggers across various platforms. Furthermore, with Meta (Facebook and Instagram) and TikTok having lax rules regarding fair use and repurposed content, food creators are now at risk of having their content stolen; more importantly, according to Smith, it lowers the viewer’s perception of the quality of the creators and potentially puts the consumer in danger with poorly created recipes.
In one of her recent videos, Smith shows her audience how to make baked macaroni and cheese. The clip starts with a blend of cheese sauce pouring over the freshly-cooked macaroni; the cheese pour has this impeccable smoothness to it, almost as if the average person could not replicate it when trying to do so themselves. Naturally, this can be credited to high-quality video production mixed with Smith’s years of expertise. Nevertheless, Smith said that dozens of people reached out to her via Instagram and TikTok DM’s, accusing her of using AI-generated content.
“These messages were just so disappointing for me,” Smith said. “It’s almost like nowadays the quality of your videos can’t be too good, or else people don’t think that it’s authentic.”
Despite Smith’s disappointment, social media users’ assumptions that what they see could be AI are justified. According to recent reports, just over half of all content online is AI-generated, with the quality of this content only getting better as the days go on. The individuals behind the AI-generated accounts are often faceless, so there is rarely someone to point out and pin the blame on.
Smith started her career with a blog she and a friend made more than 15 years ago. In the years following, she created her own blog, “By Kelsey Smith,” in 2020, and has seen a sizeable following in the half-decade since, reaching over 13,000 followers across Instagram and TikTok. To keep up with social media’s changing landscape, she moved to making videos to support her blog, and subsequently, increased the quality of those videos.
“When I first started back in 2010, quick photos without much attention to equipment, plating, lighting, or overall production quality, were the norm,” she says. “Standards have become much higher across the industry over time, with better cameras, better lighting, SEO strategy, long-form blog posts, and video production all becoming part of the job.”
Some of Smith’s most viral moments on social media actually do not come from her recipe guides. Two years ago, Smith created a series she titled “Spot the AI,” where she would educate her viewers on how to tell the differences between real versus AI-generated blog posts, recipes, and photos. The videos were short and informative; Smith did not talk down to her audience, something that is common when creators denounce AI content. She argues that at the time, her series was very successful in getting people with an untrained eye to notice the differences. The reason she stopped it, however, was that AI was becoming too advanced for her to notice the difference, and she simply could not keep up.
“I started that series to help educate people on common indicators in photography, and back then, I thought AI was fairly easy to spot to a trained eye,” Smith says. “Now, it’s often impossible to tell. And video has caught up much faster than I expected.”
AI-generated videos have seen the largest jump in both quality and frequency over the past few months, especially with the public release of Sora 2, a video generator released by the creators of ChatGPT, in late September. Prior to Sora’s release, AI videos did not have the capability to produce realistic content and were limited mostly to animations. Real-world situations are a breeze for Sora, and are nearly impossible to spot by a trained eye, let alone the average Facebook user, Jeremy Carrasco argues. Carrasco, who initially worked with media and audio encoders in the AV industry, has now amassed over 100,000 followers on TikTok for teaching users which videos are real versus AI-generated.
Carrasco is not the only person on social media who makes videos like these; he, like many others, uses his technical background to explain the differences— some of these are hidden in plain sight, he says, such as an extra prong on a fork or a “too perfect” cut on a slice of cake. What makes him stand out, Carrasco argues, is how he ties everything back to the social implications of AI videos.
“I don't think that there are many technical people who also will dive into the problems and actually have a cultural understanding of why they are problems,” Carrasco says. “So that's definitely been a hole that needed to be filled in the ecosystem here.”
Carrasco’s most viewed video since the release of Sora 2 is of him showing how impossible it can be to decipher baking content that is real versus AI-generated. The video, which has over two million views in the past two weeks, features the TikTok account @anna_dessert, a baking page based in China, that, similar to Smith, produces very smooth-looking videos of high quality, and whose comments have been flooded with AI accusations. Carrasco juxtaposes that account with @foodshow.xyz, a page that posts AI content solely in a near identical style to @anna_dessert, and the two are nearly indistinguishable; it is only after Carrasco shows multiple, hyper-detailed, and zoomed-in shots of the AI page’s videos that one could tell the difference.
“If the average person on TikTok is getting fooled by this, I would imagine that that means the majority of Americans can't figure out on first glance,” Carrasco argues.
The AI-generated videos posted by @foodshow.xyz look uncannily similar to real creators -- but what is arguably more dangerous are the recipes in the descriptions. These recipes are almost always created by AI, and often contain misleading information on how to properly cook the food. Food content creator Amanda Rettke says that, despite the AI models learning how to create safer recipes as time goes on, the number of accounts stealing content is growing exponentially.
“Unfortunately for me, but fortunately for the consumer, the AI systems are educating themselves quite consistently on best practices for cooking and things,” Rettke said.
Rettke, who is from Minnesota, started food blogging in 2008 under the name “I am Baker,” and created her second website, “I am Homesteader,” in 2016; “one sweet and one savory,” she says. Rettke has gotten her largest following on Facebook, where her two profiles for the corresponding websites have over 18 million followers. She, along with a four-person staff, comes up with and tests original recipes, along with shooting photography and videography. The sheer size of Retkke’s success, however, means that her content is perhaps stolen the most.
Rettke first noticed that her content was being plagiarized by AI in 2023 when she had stolen her recipe for a red velvet peanut butter blossom—along with altering the corresponding photo—which she originally posted all the way back in 2009.
“I noticed that it was my content because the recipe was too specific to simply be just a coincidence,” Rettke said. “And the image looked just like mine, but they doctored it a little bit, maybe to make it undetectable from copyright.”
After doing some discovery on her own, Rettke claims that there is just one entity behind the stolen content on Facebook. SFG Media Group, according to Rettke, is behind hundreds of pages on social media that steal content from food creators. The only trace that I could find about SFG Media Group was from Crunchbase, a site that provides some information about private companies; they describe SFG as a “digital media company covering trending stories, uplifting news, diy, cooking, and more.” Their website, provided by Crunchbase, ultimately proved to be a broken link.
The page Sugar & Co - Baking, which has over 220,000 followers on Facebook, is one of SFG’s accounts. They post single images of baked goods consistently throughout the day, which, for the amount of followers they have, barely get any engagement—most posts get under 50 likes and just a couple of comments. Rettke thinks that SFG is potentially buying followers in an attempt to increase its legitimacy. Each post has a corresponding link to the full recipe on the website Cooktopcove.com, where Rettke says they steal her recipes and post them in their own food blogging format. She is far from the only blogger getting her content stolen, however. The website displays hundreds of articles, and each has a byline of who the site claims to be a contributing writer or recipe specialist. The names on the bylines, however, appear to be either food bloggers or freelance food writers, with the corresponding images being taken directly from Instagram and LinkedIn. According to Rettke, the website originally would just take her content verbatim, but now uses AI to change the recipes enough so that copyright infringement can not detect them, and she believes that they are unfortunately here to stay.
“It's not going anywhere,” Rettke said. “It's not even like some scammer overseas; it's folks right here that are being taught and taking courses. They're making so much money through monetization on every platform, and it is all at our expense.”
The unpredictability behind copyright infringement cases on various social media platforms can often leave these food bloggers hopeless with regard to their content being stolen. There has been some proactiveness, however, in establishing a community to try and fight against plagiarism. Rettke started a group along with Smith and some other creators to call out the stolen AI-generated content whenever they spot it on social media, which she humbly relegates to people listening to her complain. Smith, however, said that the group has been of use to her since its inception just over two years ago.
“It's helped my AI radar and helped me stay up to date on how advanced the technology is getting,” Smith said.
On a more official note, the Food Bloggers Alliance was created earlier this year by fellow blogger Lauren LeBlanc, and their Facebook page has nearly 300 members. The goal of the alliance is to ensure that creators have some form of organized support if and when their content gets stolen. Smith is optimistic about the community, saying that the social network makes sure that everyone is informed and on the same page.
Rettke, on the other hand, holds a more pessimistic view—she thinks that the AI bubble will continue expanding no matter what because, in the end, it makes the social media platforms more profitable and that, in time, will be able to be perfect.
“I wish that I could be as good as the AI. That's what people want to see. Unfortunately, I can't do it perfectly,” she said.