AI Diet Tracking, is it ready yet?

Can you use AI to make diet tracking easy?

Diet tracking can provide powerful actionable insights, but at the cost of a significant chunk of your time. 

Can AI diet tracking make it easy for you? 

No. Unfortunately. Not yet.

This is according to a dedicated evaluation of ChatGPT-4, Claude 3.5 Sonnet and Gemini 1.5 Pro, using 52 standardised food photos including full meals and individual foods in varied portion sizes. Link in the comments.

The researchers compared each model’s estimates of weight, calories, carbs, protein and fat against reference values from weighed foods and a nutritional database, using standard validation statistics.

📜 AI Diet Tracking Key Study Findings 📜 

ChatGPT and Claude’s estimates of calories had an average error of around 36% of their true value.

For macronutrients, protein, carbohydrate and fat, estimates had a far higher average error of between 48 to 73%.

Gemini performed substantially worse.

All three models systematically underestimated larger portions, with their errors increasing as portion size increased.

​Calories and macronutrients are the easy things to estimate. Analysis did not include an estimate of micronutrient value, but if it had, it would surely have been far, far worse. 

The food photos were also simple, and well composed. Like a very distinct arrangement of a cut of meat, rice and broccoli on a plate. Replace that with a chaotic curry dish, rainbow mixed salad or a dish where components are obscured, like cottage pie, and it would again by far, far worse.

🤔 Why it matters?🤔 

Diet tracking & analysis can unlock precise, personalised enhancements to individual diet quality & the health benefits that flow from a high quality, nutrient dense diet.

Yet tracking with reasonable accuracy has a learning curve & a significant burden.

That's not all.

Tracking changes what people eat. Sometimes in a good way, by prompting reflection & awareness. Sometimes in a bad way. I’ve had clients tell me they chose not to eat a varied salad dish because of the work it would take to log it (they could have skipped the tracking of course)

Longer-term use may also raise the risk of disordered eating. 

For clinical practice, distinct from research, I believe the best solution presently is:

🥸 Up front, expert guidance on how to track

📊 Using Cronometer for the tracking (since it tracks up to 82 nutrients)

🎯 Match the tracking activity to the tracking goals

🤝 Ongoing coaching for actionable insights

On matching tracking to goals, what is of interest?

Is it energy balance or diet quality?

Is it identification of problem foods that trigger reactions?

Is it an indication of dietary pattern to enable prioritisation?

Is it to support sticking to a dietary template?

What do you use? Why do you use it? What is your biggest frustration?

---

My clients enjoy clear, specific, actionable guidance on how to use diet, supplementation, lifestyle and functional testing to reach their personal health goals and resolve their health issues.

Why not book a free call to find out how we would enable better health for you? 📲