We’ve all googled recipes from actual chefs and cooks on the internet, but have you ever tried recipes whipped up by a machine?
AI recipes, or artificial intelligence recipes, have been flooding the internet, and, as artificial as they are, we aren’t quite so sure about the intelligence bit. “A laughable cook but a serious foodie,” as a particular writer herself put it, used SideChef’s RecipeGen AI app, which turns a photo of any dish into a recipe, to cook restaurant-like meals. While the app completely misjudged the ingredients of one photo, it whipped up a complicated cooking process for the other, more complicated than the dish required. It only made her “hangry.”
It’s not surprising that most of the press up to this point about AI recipes has been not-so-great. While some argue that it doesn’t matter where the food comes from, whether an AI recipe or your grandma’s cookbook, the fact is that most of the consumer-facing LLM-based AI models are heavily flawed. They’re basically fancy probability calculators with no rationality, taste, or artistry associated with human cooking. Plus, the food doesn’t always turn out to be great, or even safe, for that matter.
What Makes A Good Recipe?
The key to creating good recipes is all in the details — and the intent of the cook or chef. There are numerous factors recipe makers consider while writing recipes: the historical connotations of the dish, how to make the techniques accessible and clear, getting the right flavour, and providing enough specifics to ensure that everything is easy to understand and follow.
So, a good recipe involves a ton of research and numerous rounds of testing and refining by someone who can adjust for clarity, texture, and taste. This process involves not just sensory organs but also heart, logic, and intuition. Can AI taste salt? I think not!
Where AI Misses The Mark
With AI-generated recipes, everyone from food bloggers and recipe developers to casual cooks is annoyed: they’re either lousy meals or life-threatening situations. To put it simply, a majority of the AI-generated recipes are part of AI internet slop.
For instance, food posts regularly going viral on Instagram or Facebook involve some mushy-looking dish, or worse, an overly gradient one, frequently generating thousands of comments and likes. However, when you dig deeper, many Facebook pages surrounding cooking now seem to be chock-full of AI-generated posts speaking to audiences almost entirely comprising bots.
The profiles and posts usually have common telltale signs suggesting they aren’t real: you’ll have recipe pictures with odd details, such as a garnish missing a shadow or a misshapen piece of meat.
However, the overarching question in this entire situation is whether these AI-generated recipes even work – and how edible or dangerous they are. While some of them might produce edible dishes, more often than not these recipes are missing the basic building blocks of making meals. Some instructions might be indecipherable or incomplete; others have poorly matched ingredients.
The worst of them all is when AI recipes mess up or completely miss the cook temperatures and times, leading to dangerously undercooked food. What’s worse is when AI recipes turn out to be dangerous. In August 2023, a Forbes report detailed how an AI recipe generator produced one for an “aromatic water mix” when a Twitter user prompted it to make a recipe with ammonia, bleach, and water. The recipe ended up producing deadly chlorine gas.
Finally, there’s also the cultural and connection factor. Chefs and cooks who run food blogs put in hours and hours of work before recipes and dishes are created. The ideas include not only discussions among themselves but also fulfilling requests for recipes from readers. Some recipes are entirely new and require a ton of experimentation and research, going out to restaurants to try out that particular dish in various versions, watching videos, or scouring the internet for ideas.
After the ideas are conceived, recipes are tested multiple times. Last but not least, they need to be signed off by multiple cooks before they’re published. It’s a labour of love.
Food Fantasies
TikTok and YouTube chefs have been staging cook-offs between AI recipes and those prepared by real chefs for years — and the “real” chefs have often prevailed. In 2022, Tasty compared two chocolate cake recipes, one developed by a professional food writer and the other generated by GPT-3. While the AI recipe was just fine, it was the food writer’s recipe that won in a blind taste test, as it was more nuanced and balanced.
Yes, AI recipes do make meal planning a whole lot easier, but there are a lot more ifs and buts involved. The algorithm might simply pick up on your “interests” after you’ve scrolled through an AI food site. While quite a few of them sound good — or at least normal — the actual instructions are less “how to make this dish” and more “here are some buzzwords that resemble this dish.” And more often than not, they’d be missing key ingredients or throwing up bizarre recipes. You need to think twice before making or tasting it, no matter the number of likes.
We’re not surprised. AI can only do so much, and perhaps it’s best to keep our taste buds – and stomachs – far, far away from it.
In case you missed:
- Can We Really Opt Out of Artificial Intelligence Online?
- The Internet’s AI Slop Problem
- The Era of Oversharing Online: Data Privacy Concerns
- Cloud Gaming: A New Era in Gaming
- Memecoins: Scheme-Coins Or Valuable Digital Assets
- Why You Should Use Password Managers in 2025
- Should Children Be Talking To AI Chatbots?
- Safe Delivery App and the NeMa Smart Bot: How AI Is Aiding Safer Births Amidst Limited Resources
- The AI Student Cheating Crisis: Who Are The Real Victims?
- Crypto Heists: How To Keep Your Cryptocurrency Safe?
