Recipe sites
How to spot an AI recipe before it ruins dinner
Some recipes online were never cooked by anyone. Here are eight warning signs, a five-minute way to check, and the habit that keeps fakes out of your kitchen.
It's 11 at night. You followed the recipe to the letter, and the thing on the baking sheet looks nothing like the photo.
You reread the steps. You check the oven. You wonder, briefly, whether you've forgotten how to cook. Then a worse thought arrives: what if nobody ever made this at all?
The dessert that was never made
That's what happened to CNN health reporter Brenda Goodman, who wrote about it in December 2025. She found a recipe on Pinterest for chocolate acorns made from peanut butter sandwich cookies. It promised a batch in 45 minutes. She's an experienced baker. She bought good chocolate.
It ended at 1 a.m., about five hours later, after an emergency grocery run, with trays of misshapen chocolate globs that looked nothing like acorns. When she went looking for the blogger, a pastry chef with a warm story about her grandmother's baking, she couldn't find her. The contact emails bounced. The same face and the same story turned up on a second site with a slightly different biography. As Goodman put it, neither the blogger nor her creations appeared to be real.
The same season, Bloomberg reporters Davey Alba and Carmen Arroyo, in a story Fortune republished, described an AI-assembled version of food blogger Eb Gargano's Christmas cake that would have people baking a 6-inch cake for 3 to 4 hours at 320 °F (160 °C). Gargano's verdict: "You'd end up with charcoal!"
Neither of those cooks did anything wrong. The recipes did.
What's actually going on
There are two different problems here, and it helps to keep them apart.
The first is fake food sites. These are blogs built quickly with AI: invented authors, generated photos, and recipes nobody tested, all wrapped in ads. Tom Critchlow of Raptive, a company that helps bloggers make money from their sites, told CNN these are "bad actors generating hundreds of thousands of brand new, fresh domains that have never been seen before." Raptive said that in 2025 it had blocked nearly 600 of the roughly 6,000 sites in its network for AI-related issues.
The economics explain why. According to Raptive, creators earn about $30 to $50 per 1,000 page views. A real food blogger might spend hours developing, testing and photographing one recipe to earn that. A fake site can earn the same from a page made in minutes.
The second is blended answers. When you search for a dish, an AI summary at the top of the page may hand you a complete recipe stitched together from several real ones. Adam Gallagher, who runs the blog Inspired Taste with his wife, Joanne, calls these "Frankenstein recipes." Bloomberg described the result as his blog's ingredients combined with instructions from other popular food blogs. Each part was tested by someone. The combination wasn't tested by anyone.
That has cost real people real money. Carrie Forrest of Clean Eating Kitchen told Bloomberg that 80% of her traffic, and her revenue, had disappeared over two years. In March 2026, after a lot of complaints from creators, Google changed recipe results in its AI Mode so you can tap a dish image and get links to the recipe sites behind it. "We've heard feedback on recipe results in AI Mode," Google's Robby Stein said, as Search Engine Journal reported, "and we're making updates to better connect people with recipe creators on the web."
None of this makes AI the villain of every kitchen. Critchlow made that point himself: a real chef using AI to tidy up their writing is a different thing. So are tools that help you organize, convert or adapt recipes you chose from people you trust. The problem is narrower and stranger: recipes that no human ever cooked, presented as if someone did.
Eight warning signs
None of these proves anything on its own. Two or three together should make you suspicious.
- The photos are too perfect. Generated food often has a strange gloss. Gallagher described people in these images as looking "a little plasticky," and food can look the same. Goodman later spotted blueberries in a loaf that were impossibly round craters, and frosting that faded straight into the cake. Real food has crumbs, drips and one ugly corner.
- The numbers don't add up. A 6-inch cake baked for four hours. A "45-minute" project that takes an evening. A yield of 24 from a mixture that would barely fill one pan. If the math feels wrong, trust that feeling.
- The steps use ingredients that aren't listed. Or the list has ingredients the steps never mention. A cook who made the dish would notice. A generator often doesn't.
- The voice has no specifics. Lots of gushing ("ridiculously cute," "honestly?") and no detail about what goes wrong. Real recipe writers tell you the chocolate will seize if it gets wet, because it happened to them.
- There's no real author. No last name, no photo that appears anywhere else, no social accounts with followers, contact emails that bounce. Goodman's blogger, "Anna," wrote in her introduction that her nephew calls her "Aunt Sarah."
- Ten thousand recipes and no person. A site that says it started in 2019 but carries a 2025 copyright, posts several recipes a day across every cuisine, and never mentions a kitchen, a family or a mistake.
- The comments don't fit. Either there are none, or they're vague praise that could sit under any recipe. Real comment sections are full of people saying they swapped the butter and what happened next.
- You've seen this recipe before. Same photo, same steps, slightly different words. Bloomberg described AI systems being used to clone bloggers' work, "lifting their photos, rewriting their recipes, and republishing the results as new 'original' work."
How to check in five minutes
You don't need to be a detective. You need five minutes and a bit of skepticism.
- Find the original creator. Search the recipe name with a distinctive phrase from it. If an AI summary gave you the recipe, click through to the sources it cites and read the real recipe there.
- Run a reverse image search. Most phone browsers can search by image. If the photo shows up first on someone else's site, years earlier, you've found the original.
- Read the comments. Look for people who actually made it, with details: "I needed ten extra minutes," "mine spread." Those are gold.
- Look for the author somewhere else. A real cook leaves a trail: a cookbook, a newsletter, a video of their hands kneading dough, an interview, years of posts. A name that exists on one website only is a warning.
Pinterest has added a setting to filter AI-generated content out of your feed, and CNN reported it came to the Food and Drinks category in December 2025. Both Gallagher and Critchlow told CNN they hadn't noticed much difference yet. Treat it as a helper, not a guarantee.
The habit that protects you
The best defense isn't spotting fakes faster. It's needing to search less.
Keep the recipes that worked, from people who clearly cooked them. Write down who made it and where you found it. Add your own notes: it needed five more minutes, the sauce wanted more lemon, the kids ate it. A recipe you've cooked and annotated beats any search result, because you're the one who tested it.
And be good to the people doing the testing. The bloggers losing traffic to fake sites are the same ones writing the long introductions people love to complain about. There's a reason for that life story at the top of the recipe, and it's mostly about paying for the kitchen. Visit their sites, buy the cookbook, leave a real comment when a recipe works. Those ads that make recipe websites crash on your phone are annoying, but they're how a real person gets paid to cook the thing before you do.
Before you cook tonight
Pull up the recipe you're planning to make this week. Check three things: is there a real author you can find somewhere else, do the ingredients match the steps, and do the comments sound like people who actually cooked it?
If it passes, cook it. Then write one line about how it went, and keep it somewhere you'll find it again.