What happens to a food photo after it is uploaded
A photo of a plate seems harmless, and usually it is. Three things around it are not: coordinates in the metadata, a person accidentally caught in the frame, and what happens to the image after it is sent.
Three places where a photo might end up
The path from pressing the button to receiving the calorie count passes through three points, each with its own rules.
| Where | What ends up there |
|---|---|
| Device | The original photo with all shooting metadata |
| Service server | The image during processing, and sometimes longer |
| AI provider | The image and request text — in cases where the service doesn't perform the recognition itself |
The third line is more surprising than the others. Very few companies train their own food recognition models: it is expensive and rarely pays off. The vast majority of services transmit the image via an API to a third-party provider, and the processing policy on that side is no longer their own.
Metadata: what the photo carries with it
Every photo from a phone contains an EXIF block: device model, date and time of capture, camera settings, and, if geotagging is enabled, coordinates accurate to within a few meters.
For a bowl of soup, this means your home address. Not an "approximate area," but the exact spot where you usually have dinner.
What to do about it:
- Check your camera settings to see if geotagging is enabled. On iOS, this is "Settings → Privacy & Security → Location Services → Camera"; on Android, it is a toggle within the camera app itself.
- Find out if the service strips metadata and where — on the device before sending or on their end. The difference is significant: in the latter case, the coordinates still travel over the network.
- Remember that messengers and social networks usually strip metadata, but sending a file directly often does not.
Who is in the frame
A food image is not personal data: it is impossible to identify a person from it. An image of a person is, and the rules for it are quite different.
Practically, this means that a snapshot of a restaurant table where faces at the next table are in the frame is no longer just a "food photo." The same applies to documents, bank cards, and screens that happen to be near the plate. Most services explicitly prohibit such uploads in their terms of use, and this is not a formality: their obligations when processing an image of a person versus an image of a salad differ radically.
A simple rule: crop to the plate. It also improves recognition — a model shown only the food provides a more accurate answer.
What to check in the privacy policy
Four questions that should have clear, written answers. If even one is missing, that in itself is an answer.
Is the image stored, and for how long? Phrases like "for the necessary period" are insufficient here. A meaningful answer is either "not stored" or a specific timeframe.
Who is it shared with? Either the provider itself or the category must be named. "Partners" is not an answer.
What goes along with the image? Email address, account ID, device ID — or none of these. A photo without identifiers and a photo tagged with your account are different objects in terms of what can be gathered from them.
Are images used for training? A separate question not covered by the answer to the first: not storing it yourself and not passing it on for training are different promises.
How it works here
To avoid sending you to a document: image analysis is performed by a third-party AI provider, selected automatically from a configured chain. Neither your email address, account ID, nor device ID is transmitted with the image — only the picture and the request text. The snapshot is not stored on our end: it resides in the server's RAM only for the duration of the processing and is not written to disk. Shooting metadata is removed on the device before transmission.
This is described in full in the "Where photos go" section of the privacy policy, including details not covered in this paragraph.
What you don't need to worry about
A separate note on a fear that is more common than it is justified. A food image by itself does not allow for identifying a person, determining their health status, or learning anything that could be used against them. A photo of a plate is just a photo of a plate.
The risk is created not by food photos, but by everything listed above: coordinates in metadata, bystanders in the frame, and the uncertainty surrounding where the image goes next. All three can be eliminated, and none require you to stop taking photos.
A note on app history
If the service shows you a feed of past analyses, at least the results are stored somewhere. It is worth finding out where exactly: in the browser on your device or on a server. This determines both who can see them and what happens when you delete your account.
You can check this in one step: open the app on another device using the same account. If the history is empty there, it is stored locally.
Frequently asked questions
- Is a photo of food considered personal data?
- Not by itself: it is impossible to identify a person from an image of a plate. A photo becomes personal data if it contains a person, a document, or other information that allows someone to be identified.
- Are coordinates shared along with the photo?
- It depends on the service and camera settings. The EXIF data of a photo with geotagging enabled contains coordinates accurate to within a few meters; some services remove metadata on the device before sending, while others do so on their end.
- Who analyzes food photos?
- As a rule, a third-party AI provider: very few services train their own recognition models. Therefore, in the privacy policy, you should look not only for the storage period but also for who the image is shared with.
Read next
- How a neural network recognizes food in a photo — A photo of a plate turns into a number on the screen in a few seconds, and from the outside, it looks like a measurement.
- Why AI confidently states incorrect figures — The unpleasant feature of machine responses is not that they can be incorrect—reference books and labels can be wrong too.
- AI nutritionist and nutrition chatbots: what can you trust — Asking a chatbot about your diet has become easier than making an appointment with a specialist, and the answer comes detailed, polite, and structured.
This article is for general information. It is not medical advice, a diagnosis, or a prescription for treatment or a diet, and it does not replace a consultation with your doctor. If you have a health condition, are pregnant, take medication, or follow a diet prescribed to you, decisions about food belong with your doctor.
Figures from regulations, guidelines and studies are given as they stood when this article was prepared and may since have changed; check them against the primary sources. This article is not advertising, an offer, or individual advice, and neither the author nor the site owner is responsible for decisions taken on the basis of it.