NYT Cooking tests AI recipe changes amid subscriber resistance

The New York Times has begun testing a chatbot-style AI tool in its Cooking app that lets subscribers modify an existing recipe for their own kitchen. The feature can suggest ingredient and equipment substitutions, adjust serving quantities and otherwise tailor instructions, while leaving the original recipe as its starting point.

The tool appears in a panel at the bottom of recipe pages. In one example reported by The Verge, a request to swap kale for baby spinach prompted advice to add the spinach later, because it wilts more quickly. Modified versions remain private and cannot be published or shared through the app.

The rollout has been gradual while the company assesses performance, response speed and the requests cooks make. The Verge reported that it was available to all NYT Cooking subscribers on iOS, with wider availability planned, while Emergent described it as a beta being made available to a limited subset of subscribers. Genevieve Ko, a senior Cooking editor, announced the tool on Oct. 25.

Ko said use of the feature was “completely optional,” but confirmed there was no setting to disable it. That distinction has become a central point of reader criticism, particularly among users who value the service as a straightforward collection of editor-tested recipes.

Nieman Lab’s analysis of 171 comments on the Times announcement, excluding 12 comments from Times staff, found about 77% negative and 14% positive. The main objections were that recipe adjustments did not require AI, that AI carries environmental costs, that subscribers did not want to pay for AI, and that automated alterations could be poor.

Some readers nevertheless said the function made dietary substitutions easier. One commenter praised its quick, apparently accurate conversion of recipes to gluten-free versions. Supporters also cited potential help with ingredient shortages, dietary restrictions and gaps in cooking experience.

The Times said Cooking staff evaluated many sample answers to bring the chatbot’s responses close to the advice its culinary experts would provide. Ko said the evaluation used the same rigor as recipe editing and was grounded in Times standards. Critics have argued that AI adaptations could diminish culinary expertise or deliver results inferior to human-tested alternatives.

The announcement’s comments were closed after four days. Jordan Cohen, a Times spokesperson, said the discussion had initially remained open so Cooking editors could answer questions and engage with readers, and that closing comments on older articles is standard practice to allow moderators to focus on newer discussions. Cohen also said many members of the Cooking community had tried the iOS tool and reported good results.

The feature extends, rather than replaces, earlier automation in the app. The Times had previously used AI to convert imperial volume measurements into metric weights, and Cooking added a separate option to halve or double a recipe’s serving size. The new tool is focused chiefly on substitutions, though it can also handle portions and equipment.

NYT Cooking contains more than 15,000 recipes from the Times archive and already lets users search by ingredient and automatically adjust serving quantities. Its new AI layer works on those curated recipes instead of generating recipes from scratch, a distinction that places the service’s editorially developed recipes at the centre of the experiment.