Updates·October 31, 2026, 01:00

Readers booed as the cooking app added more AI features

AI-generated from the sources listed, with automated source checks.

The New York Times' cooking app, NYT Cooking, has added more AI features, including suggestions for ingredient substitutions. In an analysis of 171 reader comments, about 77% were negative.

The New York Times' cooking app, NYT Cooking, has added more AI features. They include converting between volume and weight measurements, scaling recipes and, most recently, suggesting ingredient substitutions, with the cooking method adjusted accordingly. That's according to Nieman Lab.

The newest feature was announced by Genevieve Ko, senior editor and columnist at NYT Cooking, on October 23, 2026, according to Nieman Lab. AI Weekly gives a different launch date, October 29.

The feature is available to subscribers in the iOS app, and a broader rollout is planned. According to AI Weekly, it is optional and only modifies recipes written by humans. It does not create recipes from scratch.

The editorial team tested the responses

Ko writes that the editorial team spent countless hours evaluating example after example, so that the chatbot's answers come close to how the paper's culinary experts would respond.

Criticism from subscribers

Joshua Benton of Nieman Lab reviewed 171 reader comments on the announcement and excluded 12 from Times employees. About 77% of the comments were negative and 14% were positive. The rest were neutral or mixed.

The most frequent complaints were that cooking shouldn't require AI, and that AI burdens the environment through data centers. Others wrote that AI is not what they pay their subscription for. It costs roughly $390 a year. Some also felt the AI's adjustments would be bad.

A few comments were positive. One subscriber praised the AI for quickly and, so far, accurately converting recipes to gluten-free.

Many readers wanted a kill switch so the AI features can be turned off entirely. According to both Nieman Lab and AI Weekly, that is not possible at the moment.

The numbers should be taken with a grain of salt

The sentiment analysis was carried out by Benton himself using Codex, running GPT-6.1 Sol, on the collected comments. The result is his own categorization and not a representative measurement of all subscribers' views.