What EXIF data is, and what your photos are carrying

What it is

EXIF — Exchangeable Image File Format — is a specification for storing information about a photograph inside the photograph’s own file. It was designed in the 1990s so that a camera could record its settings, and it has been extended steadily since.

Structurally it is a small table of numbered fields tucked into the file’s header, before the image data begins. Software that knows to look reads it; software that does not skips over it. Either way it travels with the file, through copies, uploads, downloads and email attachments, unless something explicitly removes it.

The fields worth knowing about

GPS coordinates

The one that matters. If location services were enabled for the camera app, most phones record latitude and longitude — typically to within a few metres — along with altitude and sometimes a compass bearing.

The consequences are concrete rather than theoretical. A photograph of an item for sale, taken at home, carries your address to anybody who opens the file properly. A set of holiday photos posted while you are away records that you are away and where you live. A picture of a child in a garden records the garden.

Camera make, model and serial number

Every photograph from a given camera carries the same make and model, and many cameras also record a body serial number and a lens serial number. Those tie a set of images together as having come from one device, whoever posted them and under whatever name.

Dates and times

Usually two or three of them: when the shutter fired, when the file was created, when it was last modified. Often accurate to the second. A photograph presented as recent can be dated from its own header, and so can one presented as old.

Software and authorship

The name and version of whatever produced or edited the file, and optional artist and copyright fields. Files that have been through professional editing software often carry an XMP block as well, which can contain a surprisingly detailed edit history.

The embedded thumbnail

Many cameras store a small copy of the picture inside the file so that a preview can be shown without decoding the whole thing. This has historically been a genuine hazard: in some editors, cropping the main image left the thumbnail untouched, so the cropped-out portion remained visible in the file’s own preview.

Orientation

A single number, 1 to 8, saying which way up the picture should be displayed. It is responsible for almost every sideways photograph anybody has ever uploaded, and it has a page of its own.

Not all of it is a problem

It is worth being precise, because tools in this category tend to describe all metadata as a threat and that is not accurate.

A colour profile tells software how to interpret the file’s colours. Removing it can make a photograph look flat or oddly saturated on a wide-gamut display. The orientation tag is what keeps the photo upright. The JFIF header holds the print resolution. None of those identify anybody.

The fields that identify you are the location, the serial numbers, the dates and any authorship fields. Those are the ones to think about, and a tool that tells you which is which is more useful than one that alarms you about all of it.

When does it get stripped automatically?

Large social networks generally do remove EXIF when they process an upload. Exactly what they remove is their decision, it is not documented in detail, and it changes.

But posting on a social network is only one of the ways a photograph travels. Email attachments usually pass the original through untouched. So do many messaging apps when sending as a file rather than as a photo, most forums, most marketplaces, cloud storage links, and every job or rental application that asks you to attach a picture. Assuming the data is gone because “platforms strip it” is assuming a specific behaviour of a specific service applies everywhere.

How removal actually works

There are two approaches. One is to parse the file and delete the metadata blocks, leaving the compressed image data untouched. The other is to decode the image to pixels and write a brand new file from those pixels.

FileBelow does the second. It costs a re-encode — for a JPEG, a small and generally invisible loss at a high quality setting — and it has one significant advantage: there is no copy of the original container in the output at all, so there is no block left behind for a more thorough parser to find. The output was simply never given any metadata.

One useful side effect: because the orientation tag goes too, the correct rotation is baked into the pixels instead. The photograph then looks the same in software that reads orientation tags and software that ignores them.

Checking before and after

The reason the metadata tool shows you the fields before removing them, and reads the finished file back afterwards, is that “metadata removed” is otherwise an unverifiable claim. You could not see it before and you cannot see it afterwards, so a tool that simply asserts it has done the job is asking for trust it has not earned.

Your original file is not modified, incidentally. You get a second, clean copy; the file you already had still has everything in it. If the point of the exercise is that the original should stop existing, deleting it is a step you have to take yourself.