What happened
- Wired published on November 8, 2026 its test of NYT Cooking’s new AI tool, available in an iOS-only beta.
- The Times announced the change on October 23. Under the announcement piled up 172 almost uniformly angry comments, some threatening to cancel subscriptions, and after four days comments were turned off.
- The feature appears as a “customize this recipe” field above the ingredient list. If the user says they have no loaf pan, the tool asks which pans they do have and adjusts the recipe.
- Every change is shown in red-orange text, and the modified recipe carries a note: made with AI, not tested or endorsed by the original creators.
Why it matters
A publisher whose business is tested recipes risks its brand by letting a model rewrite them. Wired’s test shows one way to do it with guardrails: the AI is a helper, not the main act, and the system refuses when a request strays from the spirit of the dish.
The case is a reference for other newsrooms weighing similar features, because it pairs audience rejection with an execution the journalist judged competent.
The number
514 million visits in 2025. That is the figure given by the Times’ executive communications director, Jordan Cohen, for the Cooking site and app. The paper does not disclose how many subscribers the section has. Cooking’s YouTube channel has more than 1.4 million subscribers.
Context
According to NYT Cooking general manager Camilla Velasquez, the model looks first at the recipe, then at the Times’ ingredient substitution guide, then at user comments and, last, at what it calls AI “drawing on our expertise.” Recipe editors helped write the original rule set and evaluated the results. Velasquez said “you shouldn’t be chatting with” the tool.
In one example, asking to turn gochujang buttered noodles into a cookie produced a “hard redirect” to the same author’s gochujang caramel cookies. Earlier this year the Times had already used AI to scale servings and convert from imperial to metric.
Wired asked about jobs for the 60-person team. Velasquez said they are hiring “more than ever” and have no intention of publishing fewer recipes. So far users mostly rely on it for ingredient and equipment substitutions and for dietary adjustments. The reporter tested chicken and dumplings from a recipe with more than 6,000 ratings, swapped thighs for breasts and got “Chicken Breast and Dumplings,” with “breast” flagged as a change.
At Mamífero we have followed how media and platforms live with AI. We reported that Seattle Times and Newsday sued OpenAI, that Microsoft is pulling Copilot from LG and Samsung TVs, that Netflix is testing a tab that recommends by mood and that enterprise AI spending stalled between August and September.
What’s next
- The feature remains in beta and iOS-only, with no general release date.
- The Times will have to decide whether to keep the “not tested” notice and whether to open the tool to more platforms.
- Readers’ reaction suggests the reputational cost will be larger than the technical one.
Bottom line
Wired’s test does not settle whether subscribers want this, but it shows that an AI with rules written by editors and visible limits can be useful. For other publishers, the lesson is that careful execution does not replace explaining to the audience why it is being done.
Sources
- Wired, “I Used NYT Cooking’s New AI-Powered Recipe Modifier, and It’s Actually Pretty Good”, November 8, 2026, link.
- Cover image: Infrogmation of New Orleans, CC BY-SA 2.0, via Wikimedia Commons
Edited by Rodrigo Cornejo. How we select and verify: who writes these notes.
