In the spirit of Thanksgiving, I put three leading AI image generators — ChatGPT, Google Gemini, and Meta AI — to the test by prompting them to create images of seven classic Thanksgiving dishes. The goal was to assess each AI model’s ability to render these traditional foods with realism and appeal.
While none of the food I cooked up in this experiment was actually edible, some of the AI generated images looked good enough to eat, while others definitely got a hard pass. The results not only showcased the strengths and weaknesses of each platform, but also gave me a deeper appreciation for the classic Thanksgiving foods and those who spend time cooking them.
Let’s dig in on this feast for the eyes, and see what these chatbots cooked up in their attempts to capture the essences of Thanksgiving dinner.
1. Roast turkey
Prompt: “Create an image of cooked Thanksgiving turkey ready to serve.”
Google Gemini delivered a highly realistic image of a turkey that, unfortunately, looks too well done. The attention to detail including the texture of the crispy skin made this main course look exceptionally lifelike. Yet, the turkey was small and just too overdone to serve.
Meta AI generated an image that looked both real and appetizing, including juices in the meat. Plus, the tablecloth was a nice touch. However, the turkey appeared to need a bit more time in the oven to be completely ready. Also, in comparison to the ChatGPT turkey, it looked slightly smaller.
ChatGPT produced an image of a golden-brown turkey that looked well-seasoned and cooked to perfection. Everything from the color to the intricate garnish details on the serving platter resulted in an appetizing representation of the main course.
Winner: ChatGPT carved out the winning spot here.
2. Mashed potatoes with gravy
Prompt: “Create an image of a serving of mashed potatoes at Thanksgiving.”
Google Gemini rendered a photorealistic serving of mashed potatoes on what appears to be a double plate, but then the second plate sort of disappears. The hyper realistic texture of the mashed potatoes looked mouthwatering, but the gravy almost looks like caramel syrup in some places.
Meta AI presented a hearty serving of mashed potatoes with gravy, focusing on a warm presentation and a true double plate. The texture of both the mashed potatoes and the gravy were appetizing, refined, and realistic.
ChatGPT made me laugh by devoting the serving to an entire bowl – if only I could give it the win just for that! But the chatbot’s smooth, creamy mound of mashed potatoes topped with a generous pour of gravy, lacked some textural detail, and the gravy was a bit thin.
Winner: Meta AI stole the show with this one. From the presentation to the perfectly portioned potatoes, it’s clear Meta AI knows taters.
3. Green bean casserole
Prompt: “Create an image of green bean casserole for a Thanksgiving dinner.”
Google Gemini produced a very realistic-looking casserole, capturing the texture of the green beans and the crispiness of the fried onions. Yet, the image seemed to lack the inviting presentation of the other images.
Meta AI generated a depiction that combined realism with a homely feel, showcasing the casserole in a baking dish with a crispy, golden-brown topping. The presentation was inviting and detailed capturing the dish surprisingly well.
ChatGPT offered an image with vibrant green beans and a creamy sauce, topped with onion rings. The colors were vivid, giving the dish an appetizing look, though the overall appearance was somewhat too stylized and the turkey in the background looked a bit cartoonish.
Winner: Meta AI captured the essence of this side dish with just the right amount of crispiness and sauce.
4. Cranberry sauce
Prompt: “Create an image of cranberry sauce for a Thanksgiving dinner.”
Google Gemini blew me away with the realistic portrayal, highlighting the chunky texture and natural shine of the cranberries. The photographic image and presentation felt true to the actual dish.
Meta AI was very close with a well-balanced image, capturing both the vibrant color and the chunky texture of the sauce. But because the cranberries were full cranberries and not in the classic sauce, they looked slightly less appetizing and realistic.
ChatGPT created a bright, glossy bowl of cranberry sauce with a rich red hue that did not seem real. Coupled with the turkey in the background, the image was a bit too playful and artistic to come off as appetizing.
Winner: Gemini slayed the sauce on this round.
5. Stuffing
Prompt: “Create an image of Thanksgiving stuffing ready to serve.”
Google Gemini dropped the ball on this one. It produced a photorealistic image, but it doesn’t showcase the dish much at all because we can’t see it. Instead, the dark picture emphasizes the ambience rather than following the prompt.
Meta AI generated a hearty serving of stuffing, focusing on a rustic presentation with visible chunks of bread and herbs. The image was inviting, detailed and looked like it came straight out of a food magazine.
ChatGPT offered an image of stuffing with a golden-brown crust and visible herbs but the unrealistic way the stuffing is raised in some places and the chunks of celery make the whole image feel a little ‘off.’
Winner: Meta AI won for its depiction of mouthwatering presentation of stuffing.
6. Dinner rolls
Prompt: “Create Thanksgiving dinner rolls ready to serve.” Google Gemini
Meta AI generated a basket of dinner rolls with a rustic presentation, highlighting the golden crust. The rolls were just different enough from each other to look appealing and realistic, something that Gemini seemed to overlook.
ChatGPT also offered an image of a basket filled with golden-brown dinner rolls, but they were a bit too perfect. The depiction was warm and inviting, though it had an artistic quality that lacked realism once again.
Winner: Meta AI won this round with a basket of rolls that looked baked to a delightful hue and ready to serve.
7. Pumpkin pie
Prompt: “Create an image of a Thanksgiving pumpkin pie.”
Google Gemini seemed to misunderstand the assignment, because it delivered a realistic slice of pie. It captured the smooth texture of the filling and the flaky crust with a nice big dollop of whipped cream. Yet, the image seemed to go off "next day" vibes based on the presentation.
Meta AI presented a classic pumpkin pie with a golden crust and creamy filling, topped with swirls of whipped cream. The depiction was appetizing, and the bits of imperfections such as the crust crumbs on the plate made it look extra realistic.
ChatGPT created an image of a pumpkin pie with a dollop of whipped cream and cinnamon sticks. The colors were warm and inviting, though the texture appeared slightly stylized and far too artistic.
Winner: Meta AI wins for it's delicious-looking full pumpkin pie complete with visible pumpkin condensation.
Meta AI is the chatbot I would pick for all my AI generated images this holiday season. The chatbot consistently delivered the most realistic and detailed images, closely mirroring real-life appearances. ChatGPT offered more stylized, artistic representations, which, while visually appealing, leaned toward a creative interpretation that wasn’t what I was aiming for with this assignment. Google’s Gemini had several good rounds and struck a balance between realism and creativity, producing inviting images with a homely feel.
So whether you’re creating a menu, invitation, or just want a few images to get in the Thanksgiving spirit on social media, give these AI image generators a try. They each have their pros and cons, but they're worth digging into while you wait for the real meal.
More from Tom's Guide
- As a busy mom, these are the 5 best AI Storytelling apps for kids I've tried
- I tested Gemini vs ChatGPT on the iPhone with 7 prompts — here's the winner
- 5 ChatGPT prompts you can use to understand yourself better
English (US) ·