How To Tell Whether That Gorgeous Recipe On Social Media Is AI-Generated

All of Tasting Table brand's content is written and edited by real, live humans. We do not use AI chat tools to write our content.

The world of food blogging and online recipes is flooded with AI-generated content. Just how bad is the situation? Reelmind says 68% of home cooks use AI recipes at least once a week. Information made stunning by the fact that the stat and the source are essentially fictional — an AI article about how many cooks use AI recipes. However, it's not wrong to say that AI recipes are slowly becoming ubiquitous. Fortunately, there are ways to tell you're dealing with AI recipes versus those written and tested by humans.

Recipes that have unrealistic images are a sure giveaway. Some AI-generated photos are easier to spot than others, often featuring glossy or unrealistic elements. The instructions may also be vague, repetitive, or worded strangely. If the recipe glosses over important steps or uses language that is too formal or inappropriate for cooking, consider that a red flag. AI tends to hallucinate or forget things. Look for outlandish instructions, like a recipe that asks you to put horseradish in your brownies. Some AI recipes have even been deadly.

Only follow food bloggers and chefs who you can tell are real people. Even videos can be faked, but if you can see the person cooking and follow the steps from start to finish in a YouTube or TikTok video, it's likely to be real and not AI-generated. If you are looking at print-only recipes, check the website's history. Do some research: How many recipes do they post per day? Does the website seem legitimate? Are there real humans behind the content? If there's a new dish every day, or more than once a day, that could be AI.

DataIntelo reports that the AI recipe market was worth $610 million in 2024. There are services like CloudChef training AI to make Michelin-star dishes. However, a real person needs to test these recipes. The way modern AI works is by training on available information. It takes existing recipes — ones that have been tried and tested over the years — and synthesizes new ones based on those. But ... can AI write recipes?

The problem is that AI lacks an understanding of how flavors work together, as well as any practical experience with proportions or cooking times. It doesn't truly know things and can't understand taste. If it's trained on a bad recipe, it won't know that it's a bad recipe. That leaves it up to you to determine if the recipe will be good or not, and that's an extra step you shouldn't have to take. Facebook currently uses software that can scan a photo of food (which itself could be AI-generated), and then reverse-engineer a recipe from it, with ingredients and instructions based solely on the image. Countless ingredients or important steps could be overlooked.

An AI-generated recipe could, nevertheless, be perfectly fine — in fact, you can find plenty of forums online with people sharing positive experiences using them. However, you must use your best judgment, talk to people in forums, and go to trusted websites run by real people. As the technology improves, it will become harder to distinguish real recipes from AI-generated ones, but with some diligence and patience, you can still find trustworthy sources.