AI Task A – Food and Nutrition Calculator | Hind Marzouq Alahbabi Health and Wellbeing Learning E-Portfolio

AI tool and meal recorded

I used Google Gemini to estimate the nutritional content of a home-cooked meal containing chicken, rice, and a mixed vegetable salad.

The meal included a medium piece of chicken, approximately one cup of cooked rice (150–160 g), and a small salad containing cucumber, tomato, and lettuce. These were household estimates rather than weighed measurements. The exact chicken weight, preparation method, cooking oil, dressing, and added salt were not recorded.

I asked Gemini to estimate calories and key nutrients, explain uncertainty, evaluate the meal, suggest practical improvements, and identify the limitations of analysing one meal.

Medium piece of chicken
About one cup of cooked rice
Small mixed salad

Figure A1: Meal input and analysis request. The meal description, approximate portions, missing preparation details, and questions entered into Google Gemini. — **Figure A1.** Meal input and analysis request. The meal description, approximate portions, missing preparation details, and questions entered into Google Gemini.

AI-generated results

The following table reproduces Gemini’s estimates. The ranges are the tool’s estimates of preparation-related variation, not measured values or verified limits.

Nutrient	Chicken	Rice	Mixed salad	Total meal
Energy	195 kcal	200 kcal	15 kcal	410 kcal
Carbohydrates	0 g	44 g	3 g	47 g
Protein	37 g	4.2 g	0.7 g	41.9 g
Total fat	4.3 g	0.4 g	0.2 g	4.9 g
Fibre	0 g	0.6 g	1 g	1.6 g
Total sugar	0 g	0.1 g	1.8 g	1.9 g
Sodium	85 mg	2 mg	10 mg	97 mg

Energy395–530kcal

Total fat4.9–18.4g

Sodium97–850mg

Gemini reported these possible total ranges for energy, fat and sodium.

For its calculation, the tool treated the chicken as approximately 120 g of cooked, skinless chicken breast, the rice as 155 g cooked, and the salad as approximately 75 g. These were calculation choices made by the tool, rather than details I had measured.

Figure A2: Nutrient estimates and calculation details. The estimated nutrient table, totals, ranges, and the beginning of the explanation of uncertainty. — **Figure A2.** Nutrient estimates and calculation details. The estimated nutrient table, totals, ranges, and the beginning of the explanation of uncertainty.

My evaluation and accuracy check

The individual baseline figures add up to the totals shown. However, correct addition does not establish that the food estimates match my actual meal. Different chicken cuts, portion sizes, and preparation methods could change the results.

The meal contains a substantial estimated amount of protein and a carbohydrate source. It also includes vegetables, although the estimated fibre content is limited. Choosing wholegrain rice and increasing the vegetable portion would improve its variety and fibre contribution, consistent with the NHS Eatwell Guide.

I identified two issues in the AI response:

Guideline attribution

Gemini presented a 50% fruit-and-vegetable, 25% carbohydrate, and 25% protein division as consistent with the NHS Eatwell Guide. This is not the NHS guide’s division. NHS recommends that fruit and vegetables and starchy carbohydrates each make up just over a third of overall intake. It also explains that balance can be achieved across a day or week rather than at every meal.

Incomplete treatment of cooking oil

Gemini explained that oil could increase the rice’s calories and fat, but its table kept both rice values fixed. Its total ranges therefore do not consistently cover all the preparation variations described.

Gemini mentioned USDA FoodData Central, but the screenshots do not identify the exact database entries used. I therefore treat the figures as approximate AI estimates rather than independently verified nutrient values. I also would not conclude that a single-meal sodium threshold was exceeded without a defined benchmark and reliable preparation details.

Figure A3: AI evaluation of the meal. Further calculation details, the tool’s guideline comparison, strengths, and possible imbalances. The guideline attribution is critically reviewed here. — **Figure A3.** AI evaluation of the meal. Further calculation details, the tool’s guideline comparison, strengths, and possible imbalances. The guideline attribution is critically reviewed here.

Practical improvements

1
More fibre
Replace some or all of the white rice with brown rice, or add beans or lentils for more fibre.
2
More vegetables
Increase the salad portion or add cooked vegetables.
3
Better records
Record cooking oil, dressing, and salt more carefully next time. Use measured amounts rather than relying on visual estimates.

Figure A4: Suggested improvements and limitations. Gemini’s practical suggestions and explanation of why one meal cannot establish overall dietary quality. — **Figure A4.** Suggested improvements and limitations. Gemini’s practical suggestions and explanation of why one meal cannot establish overall dietary quality.

Reflection

This activity helped me look beyond whether a meal simply contains chicken, rice, and salad. Gemini estimated about 410 calories and 41.9 g of protein, but only 1.6 g of fibre. The fibre estimate made me consider the size of my salad and my choice of rice, rather than focusing only on calories.

I also learned that a detailed AI table can still contain weaknesses. The response mixed up healthy eating models and did not consistently include cooking oil in its rice estimates. This showed me why checking the explanation matters as much as reading the total.

My next steps are to increase the vegetable portion, try brown rice or add pulses, and record preparation details more carefully. These changes fit my goal of building balanced home-cooked meals without replacing familiar foods completely.

I cannot judge my whole diet from this meal. Other meals and snacks also contribute nutrients throughout the day. I would use AI as a starting point for reviewing food choices, while checking guidance and treating estimates cautiously when portions and ingredients are uncertain.