Metadata and provenance

What happens to metadata when you upload a photo

Houses, a church and winding roads sit among fields and trees.
Village from the air.Photograph by Robert Harper. Robert Harper Photograph Library, Archive.London. From the archive, shown for illustration.

Upload a carefully captioned, credited photograph to the wrong service and it comes back anonymous. Which services do this, why, and how to find out for certain takes ten minutes to learn.

When IPTC, the body that maintains the photo metadata standard, tested 15 social media sites and systems in 2016, ten of them removed at least some embedded metadata from images that were downloaded again. Creator names, copyright notices and captions went in and did not come out. The pattern it found then still describes the problem: metadata tends to survive when a file is passed along untouched, and to vanish when a platform resizes it.

This article is general information, not legal advice. Checked on 18 November 2026.

Which platforms strip metadata from photos?

The most detailed public evidence is IPTC's test from spring 2019, which uploaded marked images to 16 services and examined what came back. It recorded whether each service displayed the "four Cs" (caption, creator, copyright notice and credit line) and what remained in files obtained by "Save As" in a browser or through the service's own download button.

Photo-sharing sites, 2019

Service Displayed creator and copyright? Saved image kept Downloaded file kept
500px Yes, a few fields, correctly No save option No download
Behance No EXIF and IPTC No download
Flickr Partly EXIF and legacy IPTC; XMP stripped EXIF and IPTC
Google Photos Not correctly EXIF and XMP; legacy IPTC stripped EXIF and IPTC
Instagram No No save option No download
Pexels No Stripped Stripped
pixabay No Stripped Stripped
Unsplash No Stripped Stripped

File and messaging systems, 2019

Service Displayed creator and copyright? Saved image kept Downloaded file kept
Dropbox No Stripped EXIF and IPTC
Facebook No EXIF and legacy IPTC; XMP stripped No download
Google Drive No Not applicable EXIF and IPTC
Microsoft OneDrive No Stripped EXIF and IPTC
Twitter No Stripped No download

IPTC also tested EyeEm, fotocommunity.de and ImageSnippets. Its note on Flickr is worth quoting: "Embedded Creator replaced by account owner."

Two cautions belong with these tables. They are from 2019, and services change their software without announcement; Twitter has since been renamed X. And they describe specific routes for getting a file out. Treat them as evidence that stripping is common and uneven, not as a current guide to any one platform.

Why do platforms remove metadata?

Three reasons, only one of them deliberate.

Re-encoding. Sites make several smaller versions of each upload. The software that resizes and recompresses often writes a new file containing pixels and nothing else, unless someone has configured it to copy metadata across. IPTC's 2016 study noted that images kept at original size often kept their metadata, while rescaled ones did not. Its managing director suggested at the time that sites "might not be aware that their software strips metadata inadvertently".

File size. Metadata is small, but across billions of thumbnails it adds up, and performance tools often remove it.

Privacy. EXIF can contain GPS coordinates. Removing location data from public posts protects users who did not realise it was there. This is a good reason, but it is usually applied with a broad brush that removes the copyright notice along with the coordinates.

Which parts of the metadata survive?

A JPEG can carry three separate blocks, explained in IPTC, XMP and EXIF compared, and platforms treat them differently. The 2019 results show the split clearly: Facebook and Flickr's saved images kept EXIF and the legacy IPTC block but dropped XMP, while Google Photos kept XMP and dropped legacy IPTC.

That matters because the newer rights fields live only in XMP. The Web Statement of Rights, which Google uses for its Licensable badge, the Licensor URL and the IPTC Data Mining property, added in October 2023 to signal a position on AI training, are all XMP properties. A service that keeps "some" metadata may keep the camera model and discard the licensing link.

Signed provenance data fares no better where files are re-encoded, since Content Credentials are also stored inside the file. That technology is described in Content Credentials for photo archives.

What are the consequences for photographers and archives?

  • Uncredited pictures. A file with no creator or copyright field is easy to treat as ownerless. Each stripped copy is a potential orphan.
  • Silent opt-outs. A rights reservation embedded in the file says nothing once it is removed. This is the main weakness of metadata as an AI training signal, discussed in how to protect your photographs from AI scraping.
  • Missing search features. If your own site's image pipeline strips metadata, Google cannot read the embedded licensing fields, as the guide to the Licensable badge explains.
  • Broken context. Captions and dates are what make an archive photograph usable. Without them a picture is only a picture.

Is stripping metadata lawful?

Decided. Two statutes protect rights information, and both turn on knowledge. In the UK, section 296ZG of the Copyright, Designs and Patents Act 1988 gives a remedy against a person who knowingly and without authority removes or alters electronic rights management information, where they know or have reason to believe that this will induce, enable, facilitate or conceal an infringement. In the United States, 17 U.S.C. § 1202(b) prohibits intentionally removing copyright management information, with comparable knowledge requirements. Neither makes automated stripping during resizing unlawful as such.

Pending. No rule in the UK requires platforms to preserve metadata. The government's report on copyright and AI, published on 18 March 2026, identified metadata stripping as one of the gaps in the technical measures available to rights holders, alongside the fact that robots.txt works at the level of a site and not of a work. It mandated no standard, and said the government would work with industry on best practice. Whether that leads to anything binding is not known.

How do I check what a site does to my files?

The test takes a few minutes and settles the question for the services you actually use.

  1. Take a JPEG you do not mind publishing.
  2. Embed distinctive text: a made-up creator name, a copyright notice with today's date, a description, a licensing URL. The copyright metadata tool writes the rights fields and the captions and keywords tool the descriptive ones.
  3. Confirm the fields in the metadata viewer and keep that copy as your reference.
  4. Upload it.
  5. Retrieve it every way a visitor could: right-click and save, the download button if there is one, and the image address copied from the page.
  6. Open each retrieved file in the viewer and compare, block by block.
  7. Note the date. Repeat after the service updates.

Run the same test on your own website. Content management systems and image optimisers are frequent culprits, and many have a setting to preserve metadata that needs switching on.

What can be done about it?

Nothing makes metadata indelible, but several habits limit the damage.

  • Keep masters with full metadata and treat every published copy as expendable.
  • Put the essentials in more than one place. A credit in the visible caption text survives when the embedded credit does not.
  • Use structured data on your own pages. It sits in the HTML, so image resizing cannot remove it.
  • Add a visible mark where appropriate. Pixels survive re-encoding. A watermark deters casual reuse and identifies the source, though it can be cropped out and does not prevent copying. The watermark tool applies one in batches.
  • Record a fingerprint. A perceptual hash of each published image, made with the image hash and fingerprint tool, lets you match a stripped copy back to your original later.
  • Remove what you do not want published before you upload. Do not rely on a platform to strip location data for you. The location remover deletes GPS and keeps the rest.
  • Choose services that keep metadata when you have the choice, and tell the ones that do not.

The view from a film archive

For the Robert Harper Photograph Library the question arrives from the other direction. More than half of the archive is still on film and has yet to be scanned, so most of its photographs have no embedded metadata yet: what is known is written on sleeves, mounts and contact sheets, and has to be transcribed into each file as it is digitised. That makes it all the more important that the fields, once written, are not thrown away by the first system that handles the scan. The small selection published on Archive.London, in collections such as London, is watermarked and described in structured data on the page.

The archive is owned by 825 Ventures Ltd, the owner and rights holder. Picture researchers and agencies who want an image with its caption and credit intact can ask through the picture agencies page.

Frequently asked questions

Do social media sites strip metadata from photos?

Many do, at least partly. IPTC's tests in 2013, 2016 and 2019 found that most of the services examined removed some or all embedded metadata from images that users saved or downloaded. Results varied by platform and by whether the image had been resized. Behaviour can change at any time, so test the services you depend on.

Does Instagram remove EXIF data?

IPTC's 2019 test recorded that Instagram did not display embedded creator or copyright information and offered no save or download option through which a file's metadata could be checked. In practice, pictures shown on such platforms are re-encoded copies, and you should assume the fields you embedded are not available to people viewing them there.

Why do websites remove photo metadata?

Usually as a side effect. Sites resize and recompress uploads to save storage and bandwidth, and the software doing that work often discards metadata unless told to keep it. Some platforms also remove location data deliberately to protect users' privacy. IPTC has suggested that some sites may not realise their software strips metadata.

How can I check whether a site keeps my photo's metadata?

Embed distinctive text in the creator, copyright and description fields of a test image, upload it, then download or save the published copy and open it in a metadata viewer. Compare the EXIF, IPTC and XMP blocks with the original. Repeat for each size the site serves, since thumbnails and full-size files are often treated differently.

Not in itself. UK and US law give remedies against knowingly removing rights management information where the person knows or has reason to believe it will enable or conceal infringement. Automated stripping during resizing does not obviously meet that test. This is general information, not legal advice.

Sources

Published by Archive.London. Corrections are welcome under our editorial policy.

For picture agencies, publishers and researchers

Fashion, beauty, jewellery, music, portraits and British life from the 1960s onwards, with 52,684 analogue originals still to be scanned. Tell us what you are researching.

See representation and researchWrite to the archive