The New York Times has spent a decade persuading people to pay for recipes written by named humans with tested techniques. This week it began shipping a feature that lets a model rewrite them, and the paying audience noticed.
NYT Cooking's new tool applies modifications to existing recipes, with ingredient substitution as the headline use case: you have spinach, the recipe wants chard, the assistant tells you how the dish changes — including, in principle, that spinach wilts faster and should go in later. The feature is available to all Cooking subscribers on iOS and will be widened later. It follows earlier AI additions that convert volume to weight and scale servings up or down.
Senior editor and columnist Genevieve Ko pre-empted the objections in her announcement, naming "the impact of data centers," "the concentration of power among A.I. companies" and "the risk of A.I. escaping our control" before arguing that the tool serves "the tactile physical pleasure of cooking" and gets a "very real, very human creation" onto the table. Staff, she wrote, "spent countless hours evaluating sample after sample to get the chatbot answers to closely align with how we would answer as culinary experts," applying "the same rigor as we do to editing, grounded in Times standards."
It did not land. Nieman Lab's Joshua Benton scraped the 171 comments under the announcement and ran a sentiment analysis: roughly 77% negative, 14% positive, the rest mixed. The recurring complaints were that the feature should not require AI at all, that AI's energy use is a reason to refuse it, that subscribers pay $390 a year for human work, and that the substitutions will be worse than a cook's own judgement.
"The internet is full of AI cooking content. I pay you because I prefer mine done by people," wrote Susan from New Jersey. "This decision misses the mark by such a wide margin it's baffling," wrote Ian from Los Angeles, adding that the disclosure was equally baffling. Bill from Pittsburgh was blunter: "Nobody wants this, it's bad for cooking, it's bad for the NYT." One reader, Naomi from California, framed it as a loss: "I am sad that AI has breached the one online resource I truly love."
Dozens asked for an off switch. Ko's answer: "No, not at this time, though this feature is completely optional to use." That distinction — optional to invoke versus possible to disable — is where a lot of the irritation concentrates. An opt-in tool that sits in the interface of a product you pay for is still a change to that product.
The minority who liked it were practical. "I love the new AI feature because it quickly (and so far) accurately converts recipes to gluten free," wrote Anne Wagner of Kansas City. "I don't cook for stress relief. I cook to put dinner on the table," said Andy from San Francisco.
Both sides are describing the same design question. For a reader with a dietary restriction at 6pm on a Tuesday, a substitution engine is close to a medical accommodation. For a reader who cooks as craft, an assistant that rewrites a tested recipe reframes the expertise they are paying for as a preference.
The Times is not backing down. Nor, on the evidence, is its audience. The instructive part is not that subscribers dislike AI — plenty of consumer AI launches draw grumbling. It is that this one landed on the exact promise the subscription sells: human authorship, tested in a kitchen, by someone with a name. Media companies adding AI to paid products should read the sentiment numbers closely. Trust is the inventory. It does not regenerate.
