Metadata in digital asset management is structured data that describes a file — who created it, what it shows, how it’s built, and what you’re allowed to do with it — rather than the visual or audio content of the file itself. It’s what lets a DAM system search, filter, and govern thousands of assets without a person opening each one to check. The IPTC Photo Metadata Standard, the reference framework media organizations use to structure these fields, groups them into three categories: descriptive, administrative/technical, and rights-related.
Why “data about data” needs structure in DAM
A photo file on its own only knows its own pixels. Everything a DAM needs to make that photo findable, trustworthy, and legally usable — its subject, its format, its expiration date — has to be attached separately, as metadata. Without that layer, a library of 50,000 images is just 50,000 opaque files a person has to open one by one.
The Dublin Core Metadata Element Set, standardized in 1998 and later published as ISO 15836, is the most widely cited generic descriptive-metadata schema: 15 elements — Title, Creator, Subject, Date, Format, Rights, and others — designed to describe any resource, not just images. The IPTC Photo Metadata Standard builds on the same idea specifically for visual assets, organizing fields into descriptive, administrative, and rights-related properties.
Descriptive metadata: what the asset is
Descriptive metadata is the layer that answers “what am I looking at, and how would someone search for it?” It’s usually written or reviewed by a person, because it requires judgment a machine can only approximate.
Typical descriptive fields:
- Title — a short, human-readable name for the asset.
- Keywords/tags — searchable terms describing subject, setting, or campaign (“product shot,” “summer-2026,” “outdoor”).
- Caption/description — a sentence of context, often required for editorial or press use.
- Subject/category — where the asset fits in a taxonomy (e.g., a product line or region).
Some DAM platforms now generate a first draft of these fields with AI rather than starting from a blank field. Lyvio by Wedia, for example, uses its Smart Library block to suggest keywords and captions from image content that a human then confirms or edits — the same category of feature G2’s 2026 survey flags as central to whether AI tagging delivers value at all.
Technical and administrative metadata: how the file is built
Technical (or administrative) metadata is usually generated automatically at capture or upload, not typed by a person. It describes the file itself: format (JPEG, MP4, PDF), pixel dimensions, file size, color profile, codec, creation date, camera or software used, and a checksum or unique identifier for version tracking. A marketing team rarely edits this metadata directly — but a DAM relies on it to auto-route files (a 4K video proxy versus the master file), flag outdated formats, or block an upload that doesn’t meet a channel’s technical spec.
Rights metadata: what you’re allowed to do with it
Rights (or usage) metadata governs legal and contractual permissions: the license type, geographic restrictions, channel restrictions (web only vs. print), model or property releases on file, and — critically — an expiration date. This is the metadata category most often missing in practice, and its absence is exactly what causes a brand to keep running an ad with a photo whose license lapsed six months earlier. The IPTC’s Rights-related properties and its Extension schema exist specifically to make this trackable rather than tribal knowledge held by one person on the legal team.
| Type | Answers | Typical fields | Usually set by |
|---|---|---|---|
| Descriptive | What is this asset? | Title, keywords, caption, subject | A person (increasingly AI-assisted) |
| Technical/administrative | How is the file built? | Format, dimensions, file size, creation date | The system, automatically |
| Rights | What can I do with it? | License, restrictions, expiration date | Legal/rights holder, enforced by the system |
Why metadata quality is now an AI problem, not just a search problem
Metadata used to be mainly a findability issue: good tags meant faster search. That’s changed. According to G2’s March 2026 survey of 10 leading DAM platforms, seven of ten vendors identified consistent taxonomy and structured metadata — not the underlying AI model — as the strongest predictor of whether AI tagging and semantic search actually work for a customer, and all 10 platforms surveyed now ship some form of AI-assisted tagging or semantic search. Vendors including Bynder and 4ALLPORTAL have made the same point publicly: AI search and auto-tagging amplify whatever metadata discipline already exists — clean, consistent metadata makes AI more useful, and messy metadata makes AI confidently wrong faster than a human would be.
That’s the practical reason the three-category structure above is worth getting right early, before adding AI features on top of it, rather than retrofitting it after a library has already grown to hundreds of thousands of untagged or inconsistently tagged files.
Source:G2