There is something comforting about following a recipe that an actual person developed, cooked, tasted, and refined. Someone already made the mistakes so you hopefully do not have to, which is a pretty good reason to pay for a cooking subscription.
That expertise is central to the appeal of NYT Cooking. Now, The New York Times is adding an AI chatbot that rewrites recipes for subscribers, and I cannot help feeling that it is watering down what makes the service valuable.
In an explanation of the new features, senior editor and columnist Genevieve Ko says the customization tool is available to all Cooking subscribers through its iOS app, with broader availability planned. Users can request ingredient substitutions, change serving counts, and adapt instructions for different pans and kitchen equipment.
The chatbot incorporates those adjustments into a personalized recipe, highlighting modifications in orange. Customized versions remain private within the subscriber’s account and cannot be published or shared through Cooking.
The convenience is easy to understand. If your refrigerator has spinach instead of kale, getting a substitution worked directly into the instructions could save you from scrolling through comments while dinner gets progressively later.
This also goes beyond blindly swapping ingredient names. Ko describes an example in which the chatbot changed when spinach entered the dish because it wilts faster than kale, explaining the adjustment along the way.
Editors helped shape that behavior by evaluating sample responses, while engineers incorporated culinary guidance and Times materials, including recipes, an ingredient substitution guide, and community comments. The tool is also designed to reject unsuitable requests and recommend existing recipes when those would better meet a user’s needs.
Those are sensible precautions, but they do not turn a generated variation into a dish someone has actually prepared and tasted. The original recipe’s testing cannot automatically establish that every subsequent modification will work.
To be clear, The Times is keeping its human-developed recipes, and the chatbot is optional. Subscribers who want to follow the original instructions can continue doing so.
My objection is that a customized version still benefits from the NYT Cooking name while offering a different level of assurance. That distinction matters when the authority of the people behind a service is part of what customers are paying for.
Ko acknowledges the problem when discussing Cooking’s earlier AI feature for halving or doubling recipes. She could not promise that automatically scaled versions would work as written, but concluded that users could become better cooks by paying closer attention and relying on their senses.
Sorry, but that explanation does not entirely sit right with me. Watching the pan and checking the texture are good habits, yet presenting uncertainty in an automated feature as an educational opportunity feels a little too convenient for the company providing it.
Her assessment of the newer tool deserves careful reading, too. Ko says she feels confident in its results “99 percent of the time,” but the article supplies no published benchmark establishing a 99 percent accuracy rate.
That is an expression of confidence from a culinary professional involved in evaluating the feature. It should not be read as a measurement of how often customized recipes will succeed in subscribers’ kitchens.
The most revealing part of Ko’s explanation is her defense of recipe developers. She describes cooking as work that requires physical preparation, tasting, thoughtfulness, and soul, which is a convincing argument for the value of her colleagues.
It is also why this addition disappoints me. A recipe can carry someone’s judgment about texture, timing, flavor, and what makes a meal worth preparing; putting a chatbot in charge of revising it creates distance between that person and the home cook.
None of this makes the announcement a sudden reversal of some blanket NYT opposition to AI. Cooking had already used the technology for metric conversions and recipe scaling before introducing this broader customization tool.
There is a reasonable case for helping subscribers improvise, and nobody needs a professional test kitchen to approve every adjustment to dinner. Ko’s own example of using the tool to scale meals for a family gathering shows how it could reduce planning work and leave more time for people.
Still, convenience is only part of what a cooking service can offer. The attraction is also trusting a particular cook, learning how that person approaches food, and gradually developing your own judgment.
I would rather see NYT Cooking make that relationship stronger through accessible substitution guidance and explanations from its recipe developers. The company has plenty of human expertise worth bringing forward, and I am not convinced another chatbot makes it a better place to learn how to cook.
Support independent tech journalism
NERDS.xyz is independently owned and operated. If you enjoy my coverage of Linux, AI, hardware, cybersecurity, and tech culture, consider supporting the site on Ko-fi.
Support NERDS.xyz
