TL;DR
Get baking supplies delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
The Kitchn reports that AI-generated recipes are increasingly appearing in search results and social media, raising concerns among recipe developers about inaccurate instructions, untested methods and misattributed work. The report includes warnings from food bloggers, but it does not establish how often AI recipes fail or verify every example described.
The Kitchn has published a report warning that AI-generated and AI-assembled recipes are appearing in search results and social media, where they may look like tested instructions from a human recipe developer. The report cites food bloggers who say some recipes combine details from different sources or contain errors, but it does not quantify how common those problems are across recipe searches.
The report describes what Inspired Taste bloggers Adam and Joanne Gallagher call “Frankenstein recipes”: instructions that appear to combine ingredients or methods from separate online recipes. Adam told The Kitchn that these examples became more concerning to them after Google’s Gemini 3 arrived in November 2025. He said AI summaries were using their photos and producing combinations he considered inaccurate. Those are the bloggers’ accounts; the supplied material does not include a response from Google or an independent review of the examples.
Joanne gave their key lime pie recipe as one example. She said an AI result described the site as using three egg yolks, while its recipe uses five. She also argued that search summaries can make generated information appear to come directly from a particular publisher. The report says users may encounter AI-generated recipe content in Google search features and on TikTok, including posts from accounts whose identities are unclear.
Testing is a central concern raised by recipe developers in the story. Sally McKenney, who has published recipes on Sally’s Baking since 2011, said generated text can sound plausible without demonstrating that the method has been cooked or baked successfully. She told The Kitchn that her recipes undergo multiple rounds of testing and tasting. The article also describes visual cues raised by Kitchn staff, including food images that look unusually glossy or unevenly detailed, though appearance alone cannot establish whether a recipe was generated by AI.
Why Untested Instructions Matter
For home cooks, an unreliable recipe can mean wasted ingredients, time and money, or a dish that fails after work has already begun. Baking can be especially sensitive to quantities and method, so a believable-looking instruction is not the same as a verified one. The source does not document a food-safety incident, but readers should not assume that an AI-generated recipe has been tested or checked for safe handling.
The report also raises a concern for recipe publishers: when a search summary misstates a publisher’s method or combines material from several sites, readers may blame the original source. The Gallaghers say this can damage trust and affect the work of recipe developers. These are concerns reported by the people interviewed, rather than measured findings about the scale of economic harm.
test kitchen digital kitchen scale
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How AI Recipes Reached Search
The Kitchn places the growth of AI recipe content over the past few years. The Gallaghers told the publication they first noticed it gaining traction in March 2024. They said the issue became worse for their work after Gemini 3 appeared in November 2025, and that their awareness videos went viral in December 2025. The supplied report does not provide traffic figures, counts of affected recipes or a technical audit of search results.
AI language models generate text by predicting likely sequences rather than cooking and tasting a dish. That distinction helps explain the developers’ concern: instructions can read naturally while still containing incorrect quantities or incompatible steps. Still, an AI tool can assist a person who then tests and edits a recipe; the report’s warning is about treating generated or assembled content as reliable without knowing whether it was checked.
““They started including our photos and making ‘Frankenstein’ recipes.””
— Adam Gallagher, Inspired Taste, speaking to The Kitchn
How Often Errors Occur
The supplied report does not establish how many AI-generated recipes are inaccurate, how often search summaries misattribute recipes, or whether the examples reflect a broad pattern. It also does not include comment from Google, a systematic comparison of AI summaries with source recipes, or evidence that every recipe produced with AI is untested. A polished image or awkward account name may prompt caution, but neither proves AI involvement. The specific search features and results may also change over time.
What Cooks Can Check
The report recommends skepticism rather than treating visual or textual clues as definitive. Before cooking, readers can open the linked recipe publisher’s page, compare the ingredient list and method, and check whether the author explains testing or provides practical details. For food safety, follow reliable guidance for cooking temperatures, refrigeration and storage; do not rely on a generated instruction if it conflicts with established safety advice.
The supplied source does not announce a regulatory action, platform change or formal investigation. The immediate next step for readers is to verify recipes against a named source and use their judgment when instructions conflict. Whether search companies will change how AI recipe summaries identify their sources or indicate testing remains unclear.
Key Questions
What does “Frankenstein recipe” mean?
It is Adam Gallagher’s term for a recipe that appears to mix ingredients or directions from different online sources. The Kitchn report presents it as his description, not a formal technical category.
Does an AI-generated recipe always fail?
No. The report does not show that every AI-generated recipe is wrong. Its concern is that generated instructions may sound convincing without evidence that the recipe was tested.
How can I check whether a recipe is reliable?
Visit the publisher’s original page, compare the ingredients and steps, and look for information about the author and testing. If directions conflict or seem unsafe, use a trusted recipe and food-safety source instead.
Can an image prove a recipe was made with AI?
No. The Kitchn staff cite unusually glossy or inconsistent-looking food images as possible clues, but appearance alone cannot confirm how an image or recipe was made.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
