How accurate are Kairo's calorie numbers?
Kairo's numbers are AI estimates and can differ from the actual nutritional values. In our published benchmark of 16 June 2026, 16 of 16 test meals passed the accuracy gate; on branded items with a nutrition label the mean per-100g calorie error was about 3.5%, on generic whole foods 10–12%. In the re-run of 25 August 2026, 21, 23 and 22 of 23 cases passed across three runs; methodology, sources and raw data are on the accuracy page.
Last updated:

What did the benchmark measure?
On 16 June 2026, 16 internationally diverse meals were run through the same analysis path a real photo scan in the app uses. Every result was scored against public nutrition sources: USDA FoodData Central, Open Food Facts and published restaurant nutrition.
The result: 16 of 16 cases passed, with about 96% weighted-macro accuracy across the set. On branded items with a real nutrition label the mean per-100g error was about 3.5% (median about 2.5%); on generic whole foods it was higher, at 10–12%. Passing means: for an item with a label, the per-100g calorie estimate is within ±20% of the label.
On 25 August 2026 we measured again: the set has grown to 23 cases and was run three times in a row through the full production path, because a language model does not return the same number for the same input every time. Across the three runs, 21, 23 and 22 of 23 cases passed; the failed cases are explained one by one on the page.
All individual values, the methodology, the sources and the raw data (licence CC BY 4.0) are at How accurate is Kairo? Our accuracy benchmark.
What does the benchmark not tell you?
It is a small internal benchmark, not a comparison with other apps and not a guarantee for every single meal. Important context: every case in the re-run is a written meal description, not a photo. The set measures whether Kairo finds and computes the right nutrition values. It does not measure how good portion estimation from a photo is.
Kairo is most precise on packaged items with a real label. It is less precise on generic whole foods, where portion size and natural variation are ambiguous, and on freely estimated amounts such as dressing.
Kairo's results do not replace professional dietary advice or a medical diagnosis.
How do I make the numbers more accurate?
Name the brand or the portion size: directly in the text when you describe a meal, or afterwards for a photo via "Edit" → "Adjust by description". Packaged products are logged most accurately by barcode, because the values then come from the product database instead of an estimate.
In every meal Kairo shows its reasoning under "How I did the math" and what the values are based on under "Sources". How to change a result: Photo scan and correcting a result.
Can I set which way Kairo leans when in doubt?
Yes, with "Calorie Bias" in your profile under "My Goals". Five levels from "Underestimate More" to "Overestimate More"; "No Bias" is the most accurate estimate. The setting only applies when Kairo has to estimate, that is, when it cannot find exact nutrition data.
Related articles
- Getting started & scanning
How does scanning work and how do I correct a result?
The four ways to log a meal in Kairo: photo, description by text or voice, barcode and a photo of a menu. Plus: editing values, adjusting by description, deleting an entry.
Read the article - Getting started & scanning
What is Kairo and how do I get started?
Kairo is an AI calorie tracker for iPhone: photograph a meal, read off calories and macros. How to set up the app and run your first scan.
Read the article - Privacy & account
How does Kairo handle my data, and how do I delete my account?
Kairo is operated by Centaurio UG in Germany; database and sign-in run in an EU region. How to export your data and delete your account directly in the app.
Read the article
Didn't find your answer?
Email support at support@kairocalories.com, in English or German. Kairo is operated by Centaurio UG (haftungsbeschränkt), Hofheim am Taunus, Germany.
Common questions
Are the calories in Kairo exact?
No. They are AI estimates. How close they came in testing is published with all numbers at kairocalories.com/en/accuracy; you can correct any meal by hand.
Where is Kairo most accurate?
In the benchmark of 16 June 2026, on branded items with a nutrition label (about 3.5% mean per-100g error). On generic whole foods the error was 10–12%.
How do I report a wrong result?
Open the meal, tap the ⋯ menu and choose "Give feedback". There you can say what was wrong.

Kairo reads your meal from a photo and shows calories and macros. For iPhone, with Apple Health and Apple Watch.
Download on the App Store