Nutritionist reviews an AI diet plan

Beth Furness - Assistant Nutritionist | 14 Mar, 2024

Ever considered letting Artificial Intelligence (AI) create a diet plan for your weight loss journey? In today's digital age, technology has infiltrated every aspect of our lives, including our dietary habits. Nowadays, some people are turning to AI-generated diets to help achieve their health goals. I decided to test out this machine-learning feature and see if it passes the qualified nutritionist test!

Asking the right question
Asking the right question

Asking the right question

Asking AI to create a diet plan for me turned out to be more challenging than I initially thought! I began by requesting a "7-day healthy diet plan for weight loss". However, I quickly realised that the plan I was offered lacked quantities and weights for food, making it tricky to calculate an accurate calorie count. To address this, I then asked for the plan to include British weights and measures to get precise quantities of ingredients. Since the AI tool I used was American, it defaulted to US food/ingredient names such as 'shrimps' and 'cilantro', so it needed to be told to use British food descriptions.

All of this highlights the importance of giving very precise instructions or ‘prompts’ to the AI tool - so you need to know what to ask for. And despite my additional requests, there were still some gaps in the plan over portion sizes, plus vague descriptions like 'grilled lean steak' and 'mixed vegetables'. If you’re following a diet plan, you need exact amounts and food descriptions.

As is the case with technology in general, the output you get is only as good as the data you feed in – or in the case of AI, the prompts you give it. Make vague requests and you get poor quality output. Detail is the key. If you’re testing this approach, to make the plan personal, include instructions about any allergies or intolerances to be aware of, and food dislikes. Caution – there is no guarantee it will be correct, but the more instructions it has to work from, the better the result.

The diet plan

The quality of the diet plan

Once the diet plan had been generated, I input all the meals into the Nutracheck app to calculate a full nutritional breakdown and calorie count. Here’s what it looked like:

Calories

The calorie content per day ranged from a low 1,117 kcals to 1,617 kcals, with an average daily intake of just below 1400 kcals. The total calorie allowance for the week was around 10% under the level for a healthy rate of weight loss (1-2lbs per week) – based on a 10st, middle-aged female. This is on the lower side and potentially may be less sustainable over a long period, but for a couple of weeks wouldn’t raise undue concern. However, my big issue is that the AI diet plans I experimented with for weight loss tended to neglect calorie considerations and over time this could result in an individual having too low a calorie intake, or conversely not achieving the necessary calorie deficit to lose weight.

Nutritional balance

Types of foods: Despite the lack of detail in some areas of the diet plan, the actual foods suggested are certainly healthy and balanced, consisting of whole, unprocessed meals. My AI plan encouraged the consumption of wholegrain carbohydrates, lean proteins and unsaturated fats.

Fruit & veg: The plan promoted up to nine portions of fruits and vegetables daily – which is almost double the 5-a-day target.

Fibre: I calculated the fibre content – it worked out around an average of 26g per day (slightly below the recommended daily amount of 30g). However it's worth noting that the average person’s daily consumption in the UK is only 18g, so getting 26g would still be an improvement! In terms of encouragement to consume wholegrain carbohydrates, pulses, nuts, seeds, and plenty of fruits and vegetables, the plan is a positive step toward increasing fibre intake, which is beneficial for satiety and promoting gut health.

Salt & sugar: The content of both was also below the recommended limits which is positive.

Carbs: This one was interesting. The carbohydrate intake was a whopping 37% below the recommended target for the week, indicating a bias towards a low-carb diet - which I did not ask for. The tool frequently suggested portion sizes of starchy carbs that were even smaller than what is typically considered a ‘small portion’, which does raise concerns as wholegrain carbohydrates are good for us.

Fat & saturated fat: The content of the meals met the weekly targets for a balanced diet.

Cooking methods: The cooking methods suggested in the plan, like grilling or baking, demonstrated a good grasp of healthy food preparation - but in some instances, ingredients like 'roasted asparagus' or ‘stir-fried tofu’ lacked clarification on oil usage. Oils can significantly increase calorie content, and leaving out these details may lead to unintentional calorie overconsumption.

Is it realistic?

Cost & food preparation

The nutritional diversity offered by the AI-generated diet plan, with a variety of protein sources like fish, seafood, poultry, beef, and pulses, is great for overall health – but it comes with a price tag. Some of the ingredients recommended by the tool might be unaffordable if you’re on a tight food budget. The plan neglects the practical aspects of meal planning, where you can ensure you manage your food shop to ensure you use up ingredients across the week. And there’s also the time required for meal preparation – this isn’t specified and cooking from scratch for every meal occasion may be burdensome for individuals managing work, family, and other commitments.

What’s my verdict?

Mixed! Looking closely at the AI diet plan I generated, there are some good points, although there are downsides and I have several concerns.

Positives

  • It knew what a healthy diet should look like and the AI-generated plan included healthy, balanced meals and a variety of nutrient-rich options.
  • The meals suggested meet most recommended nutritional targets.
  • It provides a good ‘healthy diet’ template as a starting point.

Negatives

  • Missing detail on portion sizes and cooking methods, which isn’t helpful for a weight loss plan where calorie tracking is critical to achieve a calorie deficit.
  • Didn’t factor in considerations such as food preparation time and cost of ingredients.
  • Nutritional balance may be skewed depending on what is trending in the diet world.
  • Lack of personalisation.

At first glance, the results look impressive! However, it’s important to remember that AI draws from data on the internet and social platforms, so generated content may be skewed by what is trending, and not all sources will be evidence-based. This may explain the low-carb bias of my AI plan, as carb reduction has been a popular diet trend.

The quality of generated content is only as good as the prompts you give it. While it may be great at accessing a lot of data to answer a request, it doesn’t know about you, your specific likes, dislikes, allergies or intolerances. When it comes to diet plans, one size does not fit all. We each have different nutritional requirements which vary with gender, age, culture and lifestyle. And there may be gaps in the information provided, in terms of portion sizes, quantities and cooking details.

So my AI diet plan experiment concludes that it’s a great start but there are some limitations, so treat them with caution. At this stage, I believe an AI-generated diet plan should never override any personalised guidance provided by a healthcare or nutrition professional. This is good news for me as I’m not out of a job yet! However generative AI is advancing so quickly, we will keep a close eye to see what the future holds!

Nutritionist Beth Furness (ANutr), holding a BSc in Nutrition and Health, is deeply dedicated to applying evidence-based knowledge to all aspects of nutrition. Her passion lies in fostering healthy relationships with food, ensuring that everyone maintains a balanced and sustainable approach to nutrition.

This site uses cookies to personalise content and ads, provide social media features and analyse our traffic. Find out more about how we use cookies.

Choose which cookies you allow us to use. You can read more about our Cookie Policy in our Privacy Policy.