NYT Cooking tests AI recipe changes, drawing reader objections

By Sophie Lindqvist

The New York Times has made AI recipe modifications available to all NYT Cooking subscribers on the iOS app, with a wider rollout planned “soon,” senior editor Genevieve Ko said.

The feature can recommend substitutions for ingredients and cooking equipment, scale a recipe to any number of servings, and adjust instructions to a cook’s immediate needs. Ko’s stated goal is to “generally make our recipes work for what a cook needs at that moment.”

Ko announced the feature last Friday, and The New York Times is rolling it out gradually while monitoring performance. The story matters because the change already reaches every Cooking subscriber on iOS, even as readers challenge the environmental impact of AI and the limits on user control.

For The New York Times’ Cooking product, the immediate stakes are trust and control. Subscribers must decide whether the adjustments improve the recipe service enough to justify AI’s presence.

What can the AI change?

Ko said the feature can be used for specific changes: a substituted ingredient, a different pan or tool, or a new serving count. The app then produces a customized version of that recipe.

The spinach example shows how the advice goes beyond a simple ingredient swap. In the example published with Ko’s announcement, a user asks to use baby spinach instead of kale. The tool recommends the change, but also says baby spinach wilts faster and should enter the recipe at a later step.

That timing advice shows how the feature changes more than the ingredient line. It identifies the substitute, notes that baby spinach wilts faster, and moves the addition to a later step. Ko also said the feature can modify recipes around different pans and other tools, giving the cook instructions that fit the equipment at hand.

The customization has one firm boundary. The New York Times says customized recipes remain private and cannot be published or shared through the Cooking app. The feature therefore adds a layer of personal instructions without, at least for now, adding a public route for users to circulate those variations.

For now, the useful part is recipe adaptation; the firm boundary is that each adaptation stays private.

Why are readers pushing back?

Some New York Times readers objected in the comment section beneath Ko’s announcement. According to The Verge’s report, they raised the environmental impact of widespread AI adoption and questioned how AI features may change the overall NYT Cooking experience.

One reader asked whether there would be a way to disable the feature. Ko answered: “No, not at this time.” She also described the feature as “completely optional to use.” That leaves subscribers free to ignore the tool, while the answer does not give them a switch that turns it off.

That wording creates the dispute. “Optional to use” describes whether a person invokes a function. “No, not at this time” describes whether the person can remove that function from the app. Readers can object to both parts at once, especially when they value the service as a simple recipe collection.

One reader said, “Unbelievably disappointing to see,” and added: “I’ve always loved NYT Cooking and it pains me that the AI mandate has reached what used to be a wonderful and simple collection of recipes and notes from all sorts of different people.”

The comment treats the feature as a top-down change to a product people know for recipes and notes from different people, rather than as a request for another tool. That framing also explains why a practical explanation of its functions has not answered every objection. The complaint reaches beyond bad substitutions or awkward instructions to the presence of AI itself and the limits on user control.

The environmental argument broadens that disagreement beyond product design. The New York Times has presented a small, optional tool, but readers are weighing that convenience against the impact of adopting AI more widely. The announcement gives no answer to that concern.

Readers are not just testing the feature’s advice; they are testing whether its AI framing fits a product they already trust.

How does this fit Cooking’s earlier AI work?

Ko said this is not the first time The New York Times has used AI in NYT Cooking. According to Ko, the company used it two years ago to convert imperial volume measurements to metric weight measurements in recipes. Last year, the app added a feature that converts recipes to half or double the original serving size.

Ko described the serving-size control separately, rather than explicitly identifying it as another AI-powered feature. The distinction matters because it shows two different kinds of adaptation. The older tools handle units and quantities; the new tool proposes changes to ingredients, equipment, and the sequence of cooking steps.

That progression moves the technology closer to editorial judgment. Converting a measurement preserves the original recipe. Scaling a serving count changes its size but leaves its core instructions intact. Suggesting baby spinach instead of kale, then moving its addition later, changes both the content and the technique the app has published.

The history gives the current test some context. NYT Cooking already lets cooks adjust portions without leaving the app. The new release asks users to let AI make more interpretive decisions inside the recipe, where the quality and consistency of its advice matter more directly.

Cooking already adapted basic controls; the new feature now asks AI to change the recipe itself.

Why did the comment thread close?

The comments on Ko’s announcement have closed, but the New York Times says it did not cut off the exchange immediately. The announcement post stayed open for four days, with Cooking editors answering questions and addressing concerns.

NYT spokesperson Jordan Cohen told The Verge: “Comments on this article remained open for four days, with editors from Cooking taking care to answer questions, address concerns and engage directly with readers.”

He added: “We kept that conversation open to give readers a meaningful opportunity to share feedback and hear from the team. Closing comments on older articles is standard practice at The Times so that our moderators can focus on new discussions.”

Cohen also said the company sees active interest in the tool. “We’re pleased to see that many members of our Cooking community are interested in exploring this tool and are already trying it out on our iOS app; they’re reporting good results.”

His explanation addresses how The Times manages discussions on older articles. It does not settle the two specific complaints raised in the thread: the environmental impact of AI and the absence of a way to disable the feature. The statement also offers no usage figures beyond saying many members are trying it and reporting good results.

Closing an older thread answers a narrow moderation question; it leaves the substantive objections intact.

What will the wider rollout test?

The next phase is broader availability, although Ko gave no date beyond “soon.” The New York Times says it will keep monitoring the feature while it expands. That makes the current iOS release an early test rather than the finished shape of the product.

Several questions from the launch will carry into that rollout:

• Advice quality: Does the app give accurate substitutions and timing changes, as in its baby-spinach example, across different recipes?
• Privacy: Do customized recipes remain private and unshareable through the app as the rollout reaches more subscribers?
• User control: Will subscribers still have to accept the feature’s presence without a disable option, even if they never use it?
• Environmental impact: Will The New York Times address readers’ concerns about the wider adoption of AI?
• Scope: Which platforms and features come with the promised broader release?

These are not hypothetical additions to the announcement. Each follows directly from what Ko and Cohen said, what the readers asked, and what the company has not yet settled publicly. The company also has not set out the performance measures it will watch or the threshold for changing the feature.

The coming rollout will test whether making the feature optional is enough to answer readers who object to both AI’s environmental impact and its intrusion into the app.