The short answer
- Every way of estimating calories has error. In a large US study, 24-hour food recalls under-reported energy by 12 to 20% compared with measured energy use.
- Labels are averages, portions are hard to judge by eye, and photo apps can be off by almost half on a single food.
- A range (say 600–850 kcal for a meal) is more honest than a precise figure, and it’s enough to spot patterns.
- Counting isn’t harmless for everyone: among people with eating disorders, calorie apps are often seen as making things worse.
- If counting starts to cause anxiety, stop and talk to a doctor. In the UK, Beat’s helpline (England: 0808 801 0677) can help; in the US, ANAD’s helpline is (888) 375-7767.
Why estimating what you eat is so hard
Start with a humbling fact: even with methods designed by nutrition scientists, people under-report what they eat.
The OPEN study, run in 1999–2000 with 484 adults aged 40 to 69 in Maryland, compared what people reported eating with the energy they actually used, measured by doubly labelled water, the reference method. Using 24-hour recalls (an interview reconstructing everything eaten the day before), men under-reported energy by 12–14% and women by 16–20%. With food frequency questionnaires, the gap grew to 31–36% for men and 34–38% for women.
This isn’t dishonesty. People forget a cookie, a dressing, a second drink; they misjudge a portion; they recall what they “usually” eat rather than what they ate that day. If research protocols can’t escape these errors, an app on your phone won’t either.
Method 1: reading labels
It’s the easiest method for packaged food. US Nutrition Facts labels give calories per serving; UK and EU labels give kJ and kcal per 100 g or 100 ml, sometimes per portion too.
The limits:
- The number is an average. It’s calculated from the product’s average composition with standard factors (about 4 kcal per gram of carbohydrate or protein, 9 per gram of fat). In a 2010 Tufts University study, 10 supermarket frozen meals contained on average 8% more energy than stated, and 29 restaurant items 18% more.
- The label’s serving isn’t your serving. A cereal box may list a 30 g or 1-cup serving, while a bowl filled without measuring often holds much more. Checking the weight or the per-100 g figure is the only way to convert properly.
- Anything without a label escapes the maths: home cooking, markets, canteens, restaurants, family meals.
For where reference figures like “2,000 calories” come from, see how many calories you need a day.
Method 2: estimating portions
Without scales, you judge amounts by eye or with aids: spoons, cups, bowls, portion photos. This is where a lot of the error hides.
A 2020 systematic review in Nutrition Reviews catalogued 542 portion-size estimation aids (photo atlases, food models, household utensils and more). Of 334 studies, only 21 had checked an aid against the amounts people actually ate, and few of those were high quality. Image-based aids did better than food models and household utensils. In other words, “a cup” or “a handful” is convenient, but how accurate it is for you is largely unknown.
Kitchen scales remain the most accurate method at home. They’re also the most demanding, and they don’t come with you to the office or to a friend’s house.
Method 3: photos and apps
Snapping a photo of each meal means you forget less, which is real progress. But the estimate depends on who, or what, reads the photo.
When trained researchers analyse the photos, results can be good on average. In a 2012 study in Obesity, 50 adults photographed their meals for six days and the research team estimated energy from the images. Compared with doubly labelled water, the average gap was −152 kcal a day, not statistically significant. But the standard deviation was 694 kcal: fine for a group average, much less reliable for one person on one day.
When it’s consumer apps, error varies widely. A University of Sydney team tested popular apps in 2024 against sample menus built by dietitians:
- with manual logging, 16 apps overestimated a Western-style day by about 1,040 kJ on average (nearly 250 kcal) and underestimated an Asian-style day by about 1,520 kJ (over 360 kcal);
- image recognition correctly identified between 46% and 97% of food items, depending on the app;
- on single foods, the average energy error ranged from −47% to +44% across apps; for bubble milk tea, estimates were 49% to 76% too low.
Mixed dishes, sauces and cuisines under-represented in food databases fare worst. A curry, a stew or a bowl of pho doesn’t look like a steak and fries in a photo, and cooking oil is invisible.
Why a range is more honest than a number
If every estimate carries error, showing “1,987 kcal” gives a false sense of precision. A range tells the truth about what you know.
Take one day estimated meal by meal:
| Meal | Low estimate | High estimate |
|---|---|---|
| Breakfast | 350 kcal | 450 kcal |
| Lunch | 600 kcal | 850 kcal |
| Dinner | 550 kcal | 800 kcal |
| Snacks | 150 kcal | 250 kcal |
| Total | 1,650 kcal | 2,350 kcal |
Breakfast
Low estimate350 kcal
High estimate450 kcal
Lunch
Low estimate600 kcal
High estimate850 kcal
Dinner
Low estimate550 kcal
High estimate800 kcal
Snacks
Low estimate150 kcal
High estimate250 kcal
Total
Low estimate1,650 kcal
High estimate2,350 kcal
A 700 kcal spread can look huge. It simply reflects the real uncertainty of the methods above. Yet a range like this answers useful questions: do I eat roughly the same from week to week? Do sugary drinks add up? Do I skip meals on busy days?
When it’s better not to count
Counting calories isn’t neutral for everyone. A 2017 study of 493 US college students found that those who used calorie-tracking apps reported more eating concern and dietary restraint, even after accounting for body size. The same year, among 105 people being treated for an eating disorder, about 75% used a popular calorie-counting app, and 73% of those users felt it had contributed to their disorder. Both studies are cross-sectional: they show an association, not cause and effect. That’s still reason for caution.
Signs that it’s time to stop counting and talk to a professional:
- you feel guilty or anxious when you can’t log a meal;
- you turn down invitations to stay “within your numbers”;
- you skip meals or exercise to make up for eating;
- food takes up much of your thinking during the day.
Where to get help:
- UK: Beat’s helplines are free, 3pm to 8pm, Monday to Friday: England 0808 801 0677, Scotland 0808 801 0432, Wales 0808 801 0433, Northern Ireland 0808 801 0434.
- US: ANAD’s free helpline, (888) 375-7767, Monday to Friday, 9am to 9pm Central time, offers support and referrals.
Your doctor is also a good first contact. And if you’re pregnant, have diabetes or another long-term condition, any counting should be agreed with your care team.
Keeping a record without living by the numbers
For many people the point isn’t the daily total but the patterns: what you drink, meals you skip, late-night snacking. An approximate record does that job.
In Binome360, you describe a meal in a sentence or take a photo. Your assistant suggests an estimate as a range, says what makes it uncertain, and you confirm or correct it. The app doesn’t set a calorie target for you.
[photo] Dinner: chicken tikka masala with rice and a naan
Meal ready to log. The cream in the sauce and the naan change the total a lot, so the range is wide. Save it?
I only had half the naan
Updated: 800–1,150 kcal. Save it?
Nothing is saved until you confirm.
Try Binome360 for freeAnd if what you really care about is a habit (more vegetables, fewer sodas, no snacking in front of screens), track the habit instead of the calories. Our habit tracker guide explains why tracking the behaviour works better than tracking the outcome, and our plan to cut down on sugar starts with drinks.
✓ done · ★ joker (planned rest day) · empty: missed
Frequently asked questions
Are calorie counting apps accurate?
They give an order of magnitude, not a measurement. In the 2024 study of 16 apps, the average error on a sample day was over 200 kcal for each of the three menus tested, in one direction or the other depending on the cuisine. Photos help you forget less, but portions and fats stay hard to judge.
Do I need to weigh my food?
Scales are the most accurate method at home, but they’re demanding and useless away from your kitchen. To spot patterns, you don’t need them. If a health professional asks for a precise record for a medical reason, follow their method.
Is counting calories bad for you?
Not for everyone, but it isn’t risk-free either. Studies link calorie-app use with more eating concern, especially in people vulnerable to eating disorders. If counting makes you anxious or takes up a lot of headspace, stop and talk to a doctor.
How long should I count to “learn”?
No duration has been validated by research. A few days of logging is often enough to see what keeps coming up (drinks, snacks, skipped meals); beyond that, it depends on why you’re doing it. For health reasons, advice from someone who knows your situation beats any general rule.
In short
Labels, eyeballed portions and photo apps all give estimates, with measured errors in the 10–30% range and sometimes much more for a single dish. An honest range beats a falsely precise number, and tracking a habit is often more useful than tracking a total. First step: for three days, write down what you eat and drink, with no numbers at all, and see what keeps coming up.
Sources
- Amy F. Subar et al., “Using intake biomarkers to evaluate the extent of dietary misreporting in a large sample of adults: the OPEN study”, American Journal of Epidemiology, 158(1), 1–13, 2003: doi.org/10.1093/aje/kwg092.
- Lorien E. Urban et al., “The accuracy of stated energy contents of reduced-energy, commercially prepared foods”, Journal of the American Dietetic Association, 110(1), 116–123, 2010: doi.org/10.1016/j.jada.2009.10.003.
- Birdem Amoutzopoulos et al., “Portion size estimation in dietary assessment: a systematic review of existing tools, their strengths and limitations”, Nutrition Reviews, 78, 885–900, 2020: doi.org/10.1093/nutrit/nuz107.
- Corby K. Martin et al., “Validity of the Remote Food Photography Method (RFPM) for estimating energy and nutrient intake in near real-time”, Obesity, 20, 891–899, 2012: doi.org/10.1038/oby.2011.344.
- Xinyi Li, Annabelle Yin, Ha Young Choi, Virginia Chan, Margaret Allman-Farinelli and Juliana Chen, “Evaluating the Quality and Comparative Validity of Manual Food Logging and Artificial Intelligence-Enabled Food Image Recognition in Apps for Nutrition Care”, Nutrients, 16(15), 2573, 2024: pmc.ncbi.nlm.nih.gov.
- Courtney C. Simpson and Suzanne E. Mazzeo, “Calorie counting and fitness tracking technology: Associations with eating disorder symptomatology”, Eating Behaviors, 26, 89–92, 2017: doi.org/10.1016/j.eatbeh.2017.02.002.
- Cheri A. Levinson, Laura Fewell and Leigh C. Brosof, “My Fitness Pal calorie tracker usage in the eating disorders”, Eating Behaviors, 27, 14–16, 2017: doi.org/10.1016/j.eatbeh.2017.08.003.
- Beat, “Helplines”: beateatingdisorders.org.uk.
- ANAD, eating disorders helpline: anad.org.
Also available in Français.