New York Times Tests AI Recipe Customization, Stirring Subscriber Concerns Over Cooking and Technology

The New York Times is testing an AI tool that modifies recipes for individual cooking needs, including ingredient substitutions, equipment changes, and serving-size adjustments. The feature is available to all Cooking subscribers on iOS and is being rolled out gradually before a broader release, the company says.

Users can ask the tool to adapt a recipe when ingredients or kitchen tools are unavailable. In one example, substituting baby spinach for kale prompts the system to add the spinach later because it wilts faster. Customized recipes remain private and can't be published or shared through the app. The tool also handles requests such as scaling a dish to a different number of servings or adapting it to another pan.

Ingredient swaps require more than matching one food with another. Cooking time, moisture, texture, flavor, and heat tolerance all affect whether a replacement works. A sturdy green such as chard can withstand longer cooking, while spinach collapses quickly, so the timing has to change with the substitution.

That practical detail is the feature's central test. NYT Cooking senior editor Genevieve Ko said staff evaluated many examples so the chatbot's answers would align with the guidance culinary editors would give. Earlier AI features handled measurement conversions, including volume to metric weight, and adjusted recipes for half or double servings. The new system extends automation into decisions that are closer to editorial judgment and kitchen technique.

The New York Times is presenting the tool as an aid built around its existing recipes rather than as an open-ended recipe generator. Cooking staff developed and reviewed sample answers, while the company is watching performance and usage during the staged rollout. Early requests have focused heavily on ingredient substitutions.

That distinction matters because recipe instructions carry more than ingredient data. They encode timing, technique, safety, and the intended character of a dish. The Times says the feature is grounded in its editorial standards, but subscribers are still being asked to trust an automated layer that changes instructions originally reviewed by human editors. The private-only format limits distribution of modified recipes, while leaving the original editorial product intact for users who follow it as written.

Some NYT Cooking users object to the feature because they see AI as a poor fit for a product valued for human expertise and simplicity. Comments cited by The Verge raised concerns about data-center emissions, the broader effects of adding AI to consumer software, and the possibility that automated tools will crowd out the qualities that made the service appealing.

The criticism also reflects a user-control dispute. When asked whether subscribers would be able to disable the feature, Ko replied that there was no such option at the time, while emphasizing that using the tool was optional. The Times later closed comments on the announcement after 4 days. A spokesperson said editors had used that period to answer questions and engage with readers, describing the timing as standard practice.

The Cooking test places a familiar product question inside a particularly tangible setting: how much automation belongs in an activity built around judgment, touch, and sensory feedback. Ko described the goal as supporting the physical pleasure of cooking and helping users produce a real meal, while acknowledging concerns about data centers, concentrated power in the AI industry, and AI safety.

For the Times, the feature offers a way to make a subscription more useful when shoppers lack an ingredient or own different equipment. For subscribers, it raises a sharper question about whether convenience justifies another AI system inside a trusted editorial product. The dispute is therefore about product design as much as recipe accuracy. Its next test will be whether modified instructions are reliable enough to earn repeat use without weakening confidence in the human work behind the recipes.