NYT Cooking AI assistant reshapes chicken dish and prevents cake mishap in user trial
Trained on reader comments, the tool offers ingredient swaps, scaling and equipment advice while refusing choices that would spoil flavor.

On Nov.
The assistant appears as a compact text field on each recipe page, where users type prompts about ingredient changes, serving adjustments or equipment alternatives. According to the newspaper, the system draws on a large body of reader comments, enabling it to suggest realistic substitutions, convert measurements and recommend cookware without altering the original recipe, instead generating a separate version that highlights every modification.
When the tester entered the items on hand—chicken tenders, ginger, garlic, rice, broth, limes, green beans and red onions—but noted the absence of chard, cilantro and scallions, the AI quickly produced a revised recipe titled with the added green beans. It recommended a shorter browning period for the tenders, cooking the red onion alongside the chicken, and steaming the green beans over the rice during the final minutes, while marking cilantro as optional and confirming that Thai basil could serve as a suitable substitute.
Further experiments included asking for a stand‑in for makrut lime leaf, receiving a suggestion to use grated lime zest with a note about the missing floral note. The tool also converted a Dutch‑baby batter from metric to U. S. units without issue, warned that swapping cayenne for chile would add sharp heat, and rejected a request to replace ladyfingers with minced beef in a trifle. The tester noted an initial suggestion of half the required butter for the cake, which was corrected before cooking continued.
The user compared the experience with Alexa Plus, Amazon’s generative‑AI voice assistant that can pull NYT recipes through a partnership. While Alexa offered similar substitution advice, it could only relay changes verbally and occasionally introduced errors, such as suggesting an unnecessary bite‑size cut of chicken. By contrast, the NYT Cooking assistant displayed the full revised recipe on screen, making it more reliable despite requiring manual typing.
The rollout arrives as AI‑driven kitchen aides gain momentum, joining smart displays, refrigerator apps and robotic cooking devices already in homes. Editors emphasized that the assistant is programmed to decline requests likely to produce poor‑tasting results, a safeguard built from community‑sourced comment data. The tester suggested adding voice navigation to the app, which could merge the visual clarity of the NYT tool with the hands‑free convenience of voice‑only assistants, hinting at further refinements as the technology matures.
