Technology
NYT Cooking's new AI assistant makes recipe tweaks easier—but requires a careful eye
The Times' AI-powered recipe helper excels at ingredient swaps and equipment substitutions, though it still needs human oversight to avoid kitchen mishaps.
1/3
The Verge
Weeknight cooking often involves improvisation: swapping out ingredients you don't have, adjusting portions, or finding creative workarounds for missing kitchen equipment. The New York Times Cooking app has just introduced an AI assistant designed to streamline this trial-and-error process, offering real-time guidance for recipe modifications based on the publication's vast archives and community wisdom.
How it works#
The assistant appears as a text box within recipes labeled "Customize this recipe..." Users can ask questions ranging from simple ingredient substitutions to complex multi-part requests like "Scale this recipe to feed seven people, and make it vegan and gluten-free." The tool can also convert measurements, suggest equipment swaps, and highlight all modifications in a separate "Your Version" tab so the original recipe remains untouched and easily accessible.
The AI was trained on decades of NYT Cooking content—including recipes, ingredient substitution guides, and thousands of user comments—enabling it to provide context-aware suggestions rather than generic replacements. According to the Times, it's been programmed to decline requests that would produce poor results.
Real-world performance: weeknight wins#
When testing the assistant with a One-Pot Chicken and Rice recipe, the tool quickly adapted to available substitutes: chicken tenders instead of thighs, green beans replacing chard, and fresh Thai basil standing in for cilantro. The assistant noted how each swap would affect flavor and adjusted cooking times for the smaller chicken cut. The resulting dish required no extra time but delivered noticeably better results than improvising alone.
Similar tests with ingredient swaps—comparing half-and-half to cream, swapping olive varieties, or substituting spices—yielded thoughtful responses. When asked whether fresh cayenne peppers could replace dried chilies in a halibut recipe, the assistant warned that the swap would "bring significant, sharp heat." It also correctly refused an experiment to use minced beef in a sweet trifle, showing it has guardrails against absurd combinations.
The baking challenge#
More complex adjustments proved trickier. When converting a Bundt cake recipe into muffins and then halving it to yield 12 instead of 24, the assistant made smart choices—like reducing five eggs to three rather than two and a half—but the app's interface created confusion. With no way to save modified recipes separately, toggling between versions while cooking could cause dangerous mix-ups: a voice query for butter amounts returned the modified recipe's six tablespoons instead of the original's three-quarter cup, a mistake that would have yielded a dry cake.

