Fake food made by AI looks tastier than real dishes, scientists say. Can you tell which of these meals are real meals?

Some ultra-processed foods are packed with artificial colors and flavors.

But scientists have discovered that food that is 100 percent artificial looks even tastier than the real thing.

Researchers from the University of Oxford and the University of Naples Federico II discovered that people find AI-generated food images more appetizing.

And as AI images become more realistic, experts warn they could even promote unhealthy eating habits.

Can you separate the real photos from their AI copies? Check your answers at the bottom of the article to see how well you did.

2- Which of these looks better to you?

One of these images was generated using OpenAI’s DALL-E 3 – but which one could it be?

3- Juice like this is considered moderately processed

4 - While the raw ingredients are shown when eaten themselves

The researchers arranged the images according to the basic food type and showed them in three processing stages. For example: a raw carrot, carrot juice and a carrot cake

5 - Can you tell if this is real?

6 - Or does this cake look too good to be true?

Experts say the AI-generated versions can use details such as symmetry, color and shape to produce more attractive designs

The researchers showed 297 participants a series of images that were either real photos of food or AI-generated copies.

The AI ​​images were created by putting the original photos into OpenAI’s DALL-E 3 and asking the AI ​​to ‘replicate this photo’.

When the researchers showed the images side by side, the researchers found that the participants were generally very good at spotting what was created by AI and what was real.

The success rate ranged from 62 percent for the carrot juice image to 91 percent for chocolate-covered peanut images.

However, when participants were shown the images individually, they found the task much more difficult.

7 - Cooked potatoes are shown because potatoes are not eaten raw

8 - They are used to test whether humans can tell the difference between photos and AI images

When the basic ingredient is not eaten raw, the researchers used a photo in which it was simply cooked, as here with these boiled potatoes

9 - AI images are easier to recognize when they are next to the original photos

10 - It is more difficult to recognize AI images when they are displayed individually

The researchers found that people are generally very good at recognizing AI images when they are next to photos

11- The researchers found that AI-generated foods often appeared as higher in calories

12- Does this look caloric enough to be generated by an AI?

According to the scientists, AI images tend to make food appear more caloric

For images presented individually, the percentage of correctly identifying whether they were real or not dropped to as low as 26 percent in some cases.

In their article, published in Food quality and preferenceThe researchers note that while the success rate remained high overall, there were some exceptions.

They suggest this could be due to the ‘increased complexity of the task’.

13 - The AI ​​was asked to create an image based on the original photo

14 - The AI ​​images are built from scratch and are not modifications to the original

The AI ​​images were generated by asking the AI ​​to recreate the original photo. This means that they have not been modified or built from scratch

15 - Which of these two photos seems real to you?

16 - Does this look like a real photo?

The researchers say the AI ​​appears to improve the color and shape of the food it depicts

17 - Pictures of food trigger 'visual hunger' that encourages us to eat

18 - AI-generated images of high-calorie foods can lead us to make poorer food choices

The scientists say AI-generated food images could be used to promote unhealthy eating habits

The most surprising finding, however, is that people seem to find AI-generated images more attractive, especially if they don’t know they were created by AI.

The researchers asked participants to rate the images on a scale from “Not at all appetizing” to “Extremely appetizing.”

They found that when participants were told which image was AI, they generally rated both images about the same.

But when they weren’t told which image was real, participants found that the AI-generated images were significantly more palatable.

They suggest this could be due to the AI ​​using subtle changes in color, symmetry, shape and even the position of the food.

19 - The simple shape and color of an apple can make the AI ​​copy difficult to recognize

20 - Participants found these apple images most difficult to identify when presented individually

The study found that these images of apples contained the most convincing AI fakes. Can you tell which is which?

21 - When photos of food look too good, experts say we can develop unrealistic expectations

22 - These expectations can lead people to seek out less healthy, but more appealing foods

The researchers say that AI-generated food images can lead to unrealistic expectations about food

23 - The researchers suggest that food that points towards us may be perceived as less attractive

24 - Food facing away from the camera may be perceived as more attractive

The position of food affects how attractive we find it, which of these looks best based on its position

Lead author Giovanbattista Califano, from the University of Naples Federico II, said: ‘As humans, we often feel uncomfortable when objects point at us and interpret them as threats, even if they are just food.

“When the AI ​​is tasked with replicating food photos with items pointing at the viewer, such as a bunch of carrots or a piece of cake, the AI ​​often positions the food so that it is not pointing directly at the viewer.”

It could be that the AI ​​makes these kinds of subtle adjustments to the images to make them more visually appealing.

25 - The researchers say the AI ​​can use lighting to its advantage to create better-looking food

26 - Food images with poorer lighting look less appealing to the observer

Lighting also makes a significant difference in how appetizing food looks to the viewer

27 - Is this a real slice of peanut butter on toast?  Or is it AI generated

28 - AI images can be used to save costs on marketing and photography

AI image generators can now reproduce even small details, making it difficult to tell the difference

29 - The study found that these two images were the easiest to recognize as real or fake

30 - Can you tell which of these was generated by AI?

The research found that these two images were the easiest to recognize as real or fake. Can you tell which is which?

However, the researchers also warn that this could lead to problems in the future.

Because using AI to generate images is cheap, fast and easy, study supervisor Professor Charles Spence from the University of Oxford thinks this will become more common.

He said: ‘AI-generated imagery can provide cost-saving opportunities for marketers and the industry by reducing the cost of commissioning food photo shoots.’

Professor Spence also points out that these images pose a risk due to a phenomenon called ‘visual hunger’, where viewing images of food triggers appetite and cravings.

The study found that AI-generated images not only look better than the real thing, but also tend to portray foods as higher in calories.

The AI ​​tended to show more food in each image, for example more chips or larger portions of whipped cream.

If these images spread widely, Professor Spence suggests it could lead to people making less healthy choices about their food.

Professor Spence says: ‘This could potentially influence unhealthy eating behavior or promote unrealistic expectations about food among consumers.’

Can you see which images were created by AI? Check your answers here:

  1. Real
  2. Fake
  3. Real
  4. Fake
  5. Fake
  6. Real
  7. Real
  8. Fake
  9. Real
  10. Fake
  11. Fake
  12. Real
  13. Real
  14. Fake
  15. Fake
  16. Real
  17. Fake
  18. Real
  19. Real
  20. Fake
  21. Fake
  22. Real
  23. Fake
  24. Real
  25. Real
  26. Fake
  27. Real
  28. Fake
  29. Fake
  30. Real