Image Metadata Standards
Image metadata standards define where a caption, a keyword or a rights statement is stored inside a file, and which applications will find it there. Three families do most of the work: Exif for what the camera recorded, IPTC for what the picture is about and who may use it, and XMP as the container that modern tools read and write.
The practical consequence is that the same field can exist in more than one place with different values, and an export can silently keep one and drop the others. This section covers which fields survive which workflow.
Frequently asked
What is the difference between Exif, IPTC and XMP?
Exif is written by the camera and describes capture — exposure, lens, time, sometimes location. IPTC describes meaning and rights: caption, keywords, creator, usage terms. XMP is a container format that can carry both and is what most current software reads and writes. A single file frequently holds all three.
Why do my keywords disappear after editing an image?
Because the editor wrote a new file and copied only the metadata it understands. Tools differ in which blocks they preserve, and some resize or export paths drop everything except Exif. The reliable check is to inspect the exported file rather than trusting the setting, which our material covers.
Which metadata fields should a team fill in as a minimum?
Creator, caption, keywords and usage terms answer most later questions, and cost the least to fill in at ingest. Everything else is worth adding only if somebody will search on it. Optional fields that nobody maintains make a library look documented while leaving it just as hard to search.
Does metadata survive being uploaded to social platforms?
Usually not. Most consumer platforms strip metadata on upload, partly for privacy and partly for file size. That is a reason to keep the master file, with its rights information intact, inside the DAM rather than treating a published copy as the record.
Can metadata be added to files in bulk?
Yes, and it is the only realistic way to describe a legacy library. Bulk tools apply fields across selections, and the risk is applying a wrong value at scale as easily as a right one. Working on a copy and verifying a sample before committing is the standard precaution.
Where should rights and expiry information live?
In a structured field, not in a caption. Rights that live in free text cannot be reported on, which is exactly what you need when a licence is expiring across hundreds of assets. Standards define fields for usage terms precisely so this can be queried.
What is a controlled vocabulary and why does it matter?
It is a fixed list of approved terms that people select instead of typing. It matters because free text drifts — singular and plural, spelling variants, synonyms — and every variant splits a search result. Controlled terms are the single cheapest improvement to findability in most libraries.
Should metadata be stored in the file or in the database?
Both, deliberately. The database drives search and permissions; metadata written into the file survives leaving the system, which matters when an asset is delivered to someone outside it. Writing on export is the usual compromise.
How do we handle metadata for video and audio?
The same standards do not fully apply, and time-based metadata adds a dimension images do not have. Descriptive fields transfer conceptually, but where they are stored differs by container, which is why exports between video tools lose fields more often than image exports do.
Who should be allowed to add new metadata terms?
A named owner, with a request route for everyone else. Open editing produces near-duplicates within weeks; a closed list that nobody can extend produces free-text workarounds. A review step between the two is what keeps a vocabulary usable.
Reports in this section
1 report in this section.
Sources
- Category definitions and workflow descriptions on this page summarise the reviews and guides linked below. checked 2026-08-06 — PicaJet editorial method