NYT Cooking’s AI assistant saved my weeknight chicken, but almost ruined my cake

NYT Cooking’s AI Assistant: A Weeknight Helper That Almost Sabotaged My Cake

When the fridge is bare and the clock is ticking, I usually resort to the same old chicken‑and‑rice routine. This week, however, I let the NYT Cooking app’s brand‑new AI assistant take the reins. After a few trials with ingredient swaps, recipe scaling, and even a missing Bundt pan, I discovered that the tool can be a genuine time‑saver—though it still needs a human eye, especially when it tries to reinvent desserts.

---

1. How the Assistant Works

The AI lives in a modest text box on any recipe page, labeled “Customize this recipe…”. You type a request—anything from “What can I substitute for this ingredient?” to “Scale this to feed seven people, make it vegan, and convert to metric.” The assistant then generates a new version of the recipe, preserving the original under a Your Version tab. All changes are highlighted, so you can flip back at any moment.

Key capabilities include:

• Ingredient substitution (e.g., swapping cilantro for Thai basil).
• Equipment swaps (suggesting alternatives when a specific pan is unavailable).
• Measurement conversion (metric ↔ US units).
• Scaling for any number of servings, with adjustments for cooking times and proportions.

The model is trained on the app’s extensive user‑comment archive, which helps it surface practical, crowd‑tested advice while refusing obviously bad requests (it warned me against using half the butter in a cake).

---

2. Saving a Weeknight Chicken Dinner

My usual go‑to was a One‑Pot Chicken and Rice with Ginger. I only had chicken tenders, green beans, and red onions—no chard, cilantro, or scallions. I typed:

I have chicken tenders, fresh ginger, garlic, rice, broth, limes, but no chard or cilantro or scallions. I do have green beans and red onions.

Within seconds, the AI produced a new recipe titled One‑Pot Chicken and Rice with Ginger and Green Beans. It shortened the browning time for the tenders, suggested cooking the red onion alongside the chicken, and added a final eight‑minute steam for the beans. When I asked about replacing cilantro, it approved Thai basil and noted the flavor shift.

The result? A perfectly balanced dish that earned my husband’s enthusiastic approval (he even saved leftovers for lunch). Compared to hunting through comment threads for substitution ideas, the AI saved me at least ten minutes of scrolling and gave me confidence that my tweaks wouldn’t ruin the flavor.

---

3. Experimenting with Swaps and Scaling

Over the weekend I pushed the assistant with a series of more adventurous requests:

• Half‑and‑half for cream in a caramel‑apple Dutch baby – the AI confirmed the swap and adjusted the liquid ratio.
• Castelvetrano olives + smoked paprika for kalamata and chile in a roasted halibut – it warned that the heat level would change but approved the flavor profile.
• Fresh cayenne peppers for dried chile – it flagged the sharp heat increase.
• Ladyfingers for minced beef in a trifle – it correctly refused, noting the structural mismatch.

I also tested the built‑in measurement converter, which flawlessly transformed metric weights to US cups and ounces, a feature that has been part of the app for years but now integrates seamlessly with the AI workflow.

---

4. Baking: Where the AI Stumbled

Baking is where my confidence usually wanes, so I gave the assistant a real challenge: turning a Cinnamon Roll Bundt Cake into muffins—because I didn’t own a Bundt pan. The AI:

1. Adjusted the bake time from three hours to 50 minutes.
2. Split the batter into two 12‑muffin tins (24 muffins total).
3. When I asked to halve the recipe for 12 muffins, it smartly rounded five eggs to three instead of an impractical 2½.

Both the original Bundt cake (made with a borrowed pan) and the AI‑generated muffins turned out well, but the experience exposed a usability hiccup: the app does not let you save the modified version as a separate recipe. Switching between the original and AI version required scrolling back and forth, which was cumbersome with sticky fingers. Additionally, the AI initially suggested six tablespoons of butter for the muffin version—a clear over‑estimate that I caught only after double‑checking.

---

5. Voice Integration and Future Improvements

While I could type queries, a voice‑first workflow would be ideal—especially when my hands are covered in dough. My Alexa Plus on an Echo Show offers similar substitution advice but cannot display a modified recipe, and it occasionally misinterprets steps (e.g., telling me to cut chicken into bite‑size pieces when the AI never mentioned it). Integrating voice control directly into the NYT Cooking app would combine the best of both worlds: hands‑free interaction with a visual, editable recipe.

---

Conclusion

The NYT Cooking AI assistant is a practical, time‑saving tool for everyday cooking. It excels at ingredient swaps, scaling, and measurement conversion, all while preserving the integrity of the original recipe. Its training on real user comments gives it a grounded, commonsense feel that many generic AI recipe generators lack.

However, the experience also highlights two areas for growth:

1. Separate recipe saving for AI‑generated versions to avoid confusion.
2. Robust voice integration to keep hands clean during messy prep.

For the typical weeknight cook—especially those who juggle a busy schedule and a partially stocked pantry—this assistant is a welcome addition. Just keep a quick glance at the final numbers, and you’ll avoid the occasional butter‑overload mishap.