The New York Times is integrating generative AI features into its Cooking app to help users modify recipes, a move that has sparked a sharp backlash among its paying subscribers. While the tools are designed to suggest ingredient substitutions and recipe tweaks, users are voicing concerns over the reliability of AI-generated food advice and the potential erosion of the human-tested quality that defines the subscription service, as reported by Hacker News.
The Friction Between Automation and Taste
For many subscribers, the value of NYT Cooking isn’t just the list of ingredients, but the guarantee that a professional test kitchen has actually made the dish. The introduction of AI-driven modifications strikes a nerve because it introduces the possibility of “hallucinations”—instances where an AI confidently suggests an incorrect or unsafe instruction. In a culinary context, a hallucination isn’t just a factual error; it can be a safety hazard or a ruined dinner.
The tension highlights a fundamental divide in how people view AI in the kitchen.
“The internet is full of AI cooking content. I pay you because I prefer mine done by people.”
AI Errors Create Liability in Kitchen Recipes
Within the community of power users, a heated debate has emerged over the terminology used to describe AI errors. While “hallucination” is the common term, some argue that “confabulation”—the act of fabricating memories or facts—is a more accurate description of what Large Language Models (LLMs) do. Regardless of the word used, the stakes in a kitchen are higher than in a text editor. An AI that suggests an incorrect temperature for meat or a dangerous chemical substitute creates a liability that human-tested recipes traditionally eliminate.
There is also a growing skepticism regarding the necessity of generative AI for these tasks.
The Divide in User Utility
Not all subscribers share this disdain. Some find the features genuinely useful for adjusting recipes based on specific equipment or dietary needs. These users argue that AI can internalize the science of cooking more broadly than any single recipe creator. For them, the AI isn’t replacing the recipe; it’s acting as a flexible assistant that allows for a first pass at customization before the cook takes over.
However, the counter-argument is that these tasks are already easily handled by standalone AI tools. Many users claim they can use their own preferred AI models to handle substitutions, providing the AI with more personal context regarding their specific tastes and health needs than the New York Times could ever collect.
AI Tools Threaten New York Times Subscription Model
The core of the anxiety isn’t necessarily the presence of a “sparkle button” for substitutions, but what that button signals about the future of the New York Times’ business model. The fear is that today’s optional AI tool is the precursor to cutting the expensive, human-staffed test kitchen. If the “tested” nature of the recipes disappears, the primary differentiator that justifies a paid subscription vanishes.
This creates a precarious situation for the publication. If they lean too hard into AI, they risk alienating the “power users” who value authenticity and reliability. If they ignore the trend, they risk falling behind in a market where investors are pushing for AI integration to justify massive valuations. The comparison is often made to high-end dining: paying for a gourmet meal only to find out it was DoorDashed from a ghost kitchen.
As AI continues to permeate creative industries, the NYT Cooking controversy serves as a case study in the limits of automation. In fields where trust, safety, and human experience are the primary products, the “efficiency” of AI can often feel like a devaluation of the service.
