NYT Cooking bets on AI recipe tweaks — and bets readers will trust it

The New York Times is testing an AI-powered recipe customizer inside NYT Cooking, giving subscribers an iOS-only tool that can swap ingredients, change equipment, adjust diets and sometimes scale recipes. The feature is already drawing backlash from readers who worry it could weaken a brand built on carefully tested cooking guidance.

What NYT Cooking launched

NYT Cooking has introduced a beta feature that appears as a “customize this recipe” field above the ingredient list, according to WIRED’s hands-on review. The tool is limited to subscribers using iOS for now, and the Times has not said when, or whether, it will expand to other platforms.

The company’s pitch is pragmatic rather than flashy. The customizer is meant to help readers make practical changes without forcing them to start from scratch or ask a chatbot to improvise. In use, WIRED reported, it can handle ingredient substitutions, equipment changes, dietary adjustments and some requests to scale recipes up or down.

The Times says the system draws on the recipe text, the substitution guide and user comments, then uses AI-generated guidance shaped by staff expertise. Recipe editors helped build the rule set and later assessed how the feature performed, according to the reporting.

How the tool works

The customizer is designed as a guided helper, not an open-ended conversation. Camilla Velasquez, general manager of NYT Cooking, said, “You shouldn’t be chatting with it.”

That distinction matters because the product is not being positioned as a general-purpose kitchen assistant. Instead, the AI is supposed to keep users inside the boundaries of a published recipe while offering limited flexibility where a home cook might otherwise stall.

WIRED said the system also shows visible red-orange edits to the recipe, making the AI’s changes easy to see. In some cases, it can issue a “hard redirect” to a different recipe if the requested change would push the dish too far away from the original.

Why it matters

This is more than a small feature update. NYT Cooking is one of the New York Times’ most recognizable subscription products, and any AI tool added to it has to do two things at once: make cooking easier and preserve trust in recipes that readers expect to be tested, specific and reliable.

For readers, the promise is convenience. Ingredient swaps, equipment substitutions and dietary changes are among the most common adjustments people make in the kitchen, and digital recipe products have long offered basic versions of those functions. AI raises the stakes by trying to do that work faster and with more context, but it also creates a new risk: a tested recipe can be altered in ways that change its structure or quality.

For the Times, the product challenge is more delicate. NYT Cooking has built loyalty by offering recipes that feel curated and dependable, not algorithmic. If the AI feature is seen as a shortcut that muddies that identity, it could weaken the brand. If it works well, it could become a useful extension of the subscription and a model for how legacy media companies add AI without turning editorial products into chatbots.

There is also a broader industry implication. Publishers and consumer apps are racing to add AI features that appear immediately useful rather than abstractly impressive. In that environment, the key test is not whether the technology sounds advanced. It is whether users feel the product still reflects editorial judgment.

Backlash arrived quickly

The launch did not unfold quietly. WIRED reported that an announcement post dated Oct. 23 drew intense negative comments and threats from some readers to cancel subscriptions. After four days and 172 comments that were nearly all angry, commenting on the launch story was disabled.

That response underscores the emotional attachment many subscribers have to NYT Cooking. The audience is not simply buying content; it is buying confidence in the recipe itself. When a feature appears to introduce AI into that relationship, even if narrowly constrained, the reaction can be immediate and personal.

The hostility also reflects a wider unease around generative AI in consumer products. People may accept automation for sorting, summarizing or suggesting. They are less likely to embrace it where correctness is tied to taste, texture and method. A recipe that has been tuned by editors and tested in a kitchen can feel less like data and more like a promise.

How well does it work?

According to WIRED’s testing, the tool generally worked across multiple recipe modifications. That does not mean every adjustment was perfect, but the overall result was described as effective enough to be useful.

One example exposed the limits. When WIRED tested a substitution of chicken breast for chicken thighs in Alison Roman’s chicken and dumplings, the meat came out less succulent than with thighs. That is the kind of edge case that exposes the difference between a technically valid substitution and a substitution that preserves the full quality of the dish.

In other words, the feature can help users make reasonable changes, but it cannot guarantee that every variation will taste as the original recipe intended. That is an important distinction for a product being sold on the authority of a cooking brand with a reputation for carefully edited instructions.

What the feature does best

• Ingredient swaps for pantry or dietary reasons
• Equipment substitutions when a kitchen tool is unavailable
• Diet changes for users avoiding certain foods
• Some recipe scaling requests

Where it can fall short

• Substitutions that change texture or moisture in the finished dish
• Requests that drift too far from the original recipe’s structure
• Cases where a “good enough” swap is not the same as a great one

What happened after the launch

NYT Cooking has said it does not intend to publish fewer recipes because of the new tool. Camilla Velasquez told WIRED the company has a “zero intention of publishing fewer recipes,” a reassurance aimed at readers worried the AI layer might replace the editorial work that made the product valuable in the first place.

The company is also planning to let users hide the AI feature inside the app. That option appears designed to give skeptical subscribers control over what they see, a practical response to the backlash. Whether it will be enough to calm concerns is still an open question.

The Times has kept the feature in iOS-only beta and has not said when it might reach other platforms. That leaves the product in a cautious holding pattern: live enough to test, limited enough to soften the risk.

Why it matters for NYT Cooking’s business

NYT Cooking is part of the broader New York Times subscription ecosystem, and the company has every incentive to keep the product fresh without alienating the audience that pays for it. Jordan Cohen, the Times’ executive communications director, said the “site and app were visited 514 million times in 2025,” a reminder of the scale of the company’s digital audience and the value of its product mix.

The Times also has a large audience beyond the recipe product itself, including more than 1.4 million YouTube subscribers, showing how important its digital distribution channels have become. That reach matters because product changes in one corner of the company can reverberate across the broader brand.

For NYT Cooking, the question is not whether the audience is large enough to absorb a new feature. It is whether the audience sees the feature as an aid to the brand or a compromise of it. That distinction will likely determine how much room the company has to expand AI deeper into cooking content.

Background: how the Times got here

NYT Cooking is the New York Times’ recipe and cooking-content subscription product. It sits at the intersection of journalism, service content and paid digital media, which makes it an ideal place to test tools that promise practical help.

Recipe scaling, ingredient substitution and measurement conversion are already standard digital cooking features, so the idea of a more flexible recipe helper is not a leap in concept. What changes with AI is the level of adaptation. Instead of merely converting quantities or offering a substitution list, the tool can reframe a recipe in response to a user’s needs.

That promise comes with risk. AI recipe tools can change ingredient lists and instructions, but doing so can undermine the careful balance that makes a recipe work. Unlike a news summary or a playlist recommendation, a cooking instruction has a physical end point: the finished dish. If the substitutions are off, the error shows up at the table.

NYT Cooking’s audience may be especially sensitive to that problem. The product has long been associated with editorial voices and recognizable contributors, and readers often engage heavily with comments and video personalities. That habit of participation can make product changes feel immediate and personal.

How does the new AI recipe customizer compare with traditional tools?

The simplest answer is that it is more ambitious than a converter and less open-ended than a chatbot. Traditional cooking tools typically help with numbers: how much to buy, how to scale a recipe, how to translate measurements. The Times’ new feature goes further by interpreting what kind of substitution or adjustment makes sense for a specific recipe.

That means the system is not merely calculating. It is making recommendations, and recommendations carry editorial weight. The Times appears to be trying to preserve that weight by grounding the feature in recipe text, substitution guidance and human review rather than letting it roam freely.

Still, the product sits in a gray zone. It is AI, but it is not supposed to behave like a chat assistant. It changes recipes, but it is not meant to replace them. It helps users customize, but it also has to tell them when a requested adjustment is a bad idea.

What happens next?

The biggest near-term question is whether the AI feature stays confined to iOS or eventually expands to other platforms. The Times has not offered a timetable.

Another question is whether the hide-feature option will meaningfully reduce anger from subscribers who dislike the concept of AI in NYT Cooking. Hiding the tool may satisfy some users, but it will not answer the larger concern about what AI means for recipe integrity.

There is also a product question inside the product question: how far will the Times allow the AI to go in making substitutions? If the system becomes too cautious, it may feel underpowered. If it becomes too permissive, it may produce changes that undermine recipes.

The final issue is adoption. The Times will need to learn whether subscribers actually use the customizer frequently enough to justify its place in the app, or whether most people will keep reaching for traditional recipes and editorial guidance. That will help determine whether this is a foundational feature or a narrow experiment.

By the numbers

Some of the clearest signals around this launch are numerical rather than editorial. They show the scale of the audience, the intensity of the reaction and the limited size of the beta test.

Bottom line

NYT Cooking is trying to use AI as a practical kitchen helper without turning its recipes into an open-ended chatbot experience. The tool appears capable of useful substitutions and adjustments, but the backlash shows that readers are watching closely for any sign that the Times is loosening the standards that made the product successful in the first place.

The company’s next move will be measured less by the novelty of the feature than by whether subscribers trust it to stay within the spirit of the recipe.