Ai-generated recipes are flooding tiktok and the results taste like failure
Scroll far enough on TikTok and you’ll find a robot voice promising golden, twisted cheese buns that crunch like a Dorito yet pull apart like brioche. Make them and you end up with a rubbery omelet that smells like feet. Welcome to the latest internet delusion: synthetic cooking instructions that look Michelin but eat like cardboard.
When the algorithm writes the menu, salt can’t save you
Mia Mercado, tech columnist at The Cut, spent a week replicating the platform’s most viral AI recipes so the rest of us don’t have to. First victim: cottage-cheese twists whose on-screen dough defied gravity—glossy, elastic, perfectly coiled. In her Manhattan kitchen the same batter pooled like pancake soup. “Three strips dissolved the moment I lifted them,” she told me. “What remained tasted like a confused tortilla.” Rating: zero out of five.
She is not an outlier. Search #AIRecipe on TikTok and you’ll hit 190 million views of neon-lit skillets, voice-overs stripped of human breath, and comment sections littered with photos of gray sludge. The formula is always identical: an LLM scrapes real food blogs, swaps adjectives for superlatives, slaps on a cinematic soundtrack, and watches the engagement roll in. No test kitchen, no liability, no palate.
Plagiarism is the secret spice
Mercado’s second attempt—spicy buffalo-chickpea wraps—actually tasted decent. Then she ran the transcript through a similarity checker: entire sentences lifted verbatim from Minimalist Baker, a site that spends money on groceries and human tasting. Credit? Nowhere. The AI had laundered the recipe through a thesaurus and claimed originality. Meanwhile the creator economy’s middlemen pocket sponsorships for eyeballs, not ethics.
Italian nonnas entered the chat when fettuccine in pineapple-cashew cream dropped. One non-Italian bot decided tropical pasta belonged in the canon. Mercado’s verdict: “Imagine Alfredo on a piña-colada bender, then curdle it with rage.” She scored it 1/5, charity point for edible noodles.
Food culture is becoming a low-cost training set
The damage isn’t just gastronomic; it’s archival. Every time a platform rewards synthetic slop, human recipe writers—those who sweat over hydration ratios and cultural footnotes—sink lower in the feed. Ad revenue migrates to accounts that never buy butter. Meanwhile home cooks waste ingredients during the worst food-price inflation in 40 years. The USDA estimates the average American will spend 8 % more on groceries this year; AI chefs spend nothing, because they do not chew.
TikTok’s policy team says it labels AI-generated media “when appropriate.” The tag is optional, buried, and no one uses it. Enforcement is a shrug emoji wearing a terms-of-service hoodie.
The only fix is a tougher audience
Mercado has quit the experiment. She deleted the saved folder of bot recipes and returned to cookbooks with cracked spines and marginalia in actual ink. Her parting advice: if the voice never inhales, if the hands never burn, if the garlic floats upward in slow-motion CGI, close the app and open the fridge. Taste still originates in neurons, not parameters. The algorithm can fake sizzle, but it can’t fake the sound of a real crust shattering between your teeth—an ASMR no model has trained on yet.