When The New York Times introduced an AI-powered tool for its Cooking app, the reaction from its devoted community was swift and largely negative. Yet after testing the beta feature, which is currently limited to iOS, it becomes clear that the NYT Cooking AI recipe modifier is not the kitchen apocalypse some feared. It is, in fact, a surprisingly capable assistant that mostly does what it promises: helping home cooks adapt recipes to their circumstances without upending the original spirit of the dish.
What the Tool Does
The feature appears as a “customize this recipe” field above the ingredient list. Subscribers can type requests—swapping ground pork for ground beef, replacing a loaf pan with a different pan, or adjusting for dietary needs—and the AI modifies the recipe accordingly. Any changes are highlighted in red-orange text, and users can toggle between the original and the modified version. The system draws on the recipe itself, the Times’ ingredient substitution guide, user comments, and what the company calls its own expertise, in that order.
Camilla Velasquez, general manager of NYT Cooking, describes the AI as a helper rather than the main act. She notes that the team put significant effort into telling users when a request falls outside the spirit of a recipe. For instance, asking to turn a gochujang buttered noodle dish into a cookie triggers a “hard redirect” to a more appropriate recipe—in that case, gochujang caramel cookies.
Testing the Feature
Over several days of cooking, the tool handled a range of requests. It successfully converted a chicken and dumplings recipe from bone-in thighs to boneless breasts, adjusting simmering times and adding butter to compensate for less rendered fat. It also transformed a roasted mushroom and broccoli grain bowl into a version with cauliflower and delicata squash, preserving the original recipe’s poetic voice. A waffle recipe became a yeasted version with an overnight rest, and a pasta and chickpea stew accepted extra vegetables with sensible placement in the method.
Most modifications were integrated smoothly. The AI even instructed a user to cook added fish first, then wipe out the pan before cooking tofu—a detail that showed attention to proper technique. For a sauerkraut recipe, it suggested grating carrots, slicing garlic, and mincing kale, along with an adjusted salt quantity.
One Notable Flaw
The largest issue arose with the chicken and dumplings conversion. The AI suggested returning shredded cooked breast meat to the pot for another 30 minutes while the sauce thickened and dumplings cooked. That extra half hour left the meat noticeably less succulent than pieces reserved and added later. It was a rare misstep in an otherwise solid performance, but one that highlights the limits of algorithmic recipe adaptation.
Reaction and Response
The feature’s launch on October 23 drew 172 mostly angry comments before the Times disabled commenting on the announcement. Users threatened to cancel subscriptions and questioned the need for AI in a section known for human expertise. Velasquez says the team has been hiring more than ever and has no intention of publishing fewer recipes. She also confirmed that users will be able to hide the AI feature if they prefer.
For now, the tool remains in iOS-only beta, with no announced timeline for other platforms. The concern from commenters seems rooted in a fear of losing the human connection that defines NYT Cooking’s community. The AI, at its best, saves a trip to the store or helps use up pantry items—an incremental improvement rather than a replacement for the recipe developers who built the section’s reputation.
