NYT Cooking AI Assistant Helps Weeknight Dinners, Complicates Baking

The New York Times has added an artificial intelligence assistant to its NYT Cooking subscription service, turning the popular recipe app into an interactive cooking tool that can adapt dishes in real time. The feature, available to subscribers on iOS at launch, appears as a text box on recipe pages and invites users to customize instructions. It can answer substitution questions, scale recipes for different serving sizes, adjust dietary requirements such as vegan or gluten-free cooking, suggest equipment alternatives and convert measurements. Rather than changing the source recipe, the assistant generates a new version with a separate tab and highlights altered ingredients and steps. According to The New York Times, the tool was trained in part on NYT Cooking’s extensive user comment sections, where readers have long discussed ingredient swaps, technique fixes and recipe adjustments. In hands-on testing, the assistant offered practical advice for common kitchen problems, including replacing missing herbs, adjusting cooking times for different cuts of chicken and adapting a Bundt cake recipe for muffins. The tool proved most useful as a sous chef for everyday improvisation. In one case, it rewrote a one-pot chicken and rice recipe to use chicken tenders, green beans and red onions, producing a workable dish while preserving the original for reference. It also correctly warned against inappropriate substitutions, such as using minced beef in a dessert trifle, and flagged flavor risks when swapping spicy ingredients. The assistant’s advice overlapped with Amazon’s Alexa Plus on a smart display, though the Alexa version could not display a modified recipe and became less reliable when asked to narrate changes. Baking exposed the limits of the tool. When a cake recipe was converted to muffins and then scaled down, the assistant made generally sensible adjustments, including converting metric weights and dividing batter. But switching between the original and modified recipes caused confusion, and the assistant initially provided the wrong butter amount from the alternate version. That mistake, which could have produced a dry cake, underscores why the feature still requires close human oversight. NYT Cooking users may respond differently depending on how they use the service. Recipe purists may view customized versions as a dilution of tested recipes, while casual cooks may appreciate faster access to the kind of practical guidance often found buried in reader comments. The assistant does not replace careful cooking judgment, but it offers a streamlined way to adapt recipes when ingredients, equipment or household needs differ from those assumed by the original instructions. For subscribers willing to verify amounts and cooking times, the feature represents a practical, narrowly applied use of AI in food media.