The IPTC standard defines dozens of properties and structures. A film archive needs about eighteen fields filled properly, and five of those do most of the commercial work.
A scan with nothing embedded in it becomes an orphan the moment it leaves the building. The picture may be superb, but the file cannot say who made it, who owns it or where a licence can be had, and the person who finds it three reposts later has no means of asking. The IPTC Photo Metadata Standard exists to prevent exactly that. The practical question for an archive is which of its many properties deserve the cataloguer's time.
The answer is fewer than most people fear. The standard, now at version 2025.1 (released on 27 November 2025), defines dozens of properties and nested structures across its Core and Extension schemas. A working archive can do nearly everything it needs with about eighteen, and you can see which of them any given file already carries by dropping it into the free EXIF and metadata viewer, which reads the file in your browser without uploading it.
What is IPTC photo metadata, and how does it differ from EXIF and XMP?
Three terms get used interchangeably and should not be.
Exif is the technical record written at capture: device, exposure, pixel dimensions, capture time, sometimes GPS. For a scanned transparency it describes the scanner, not the photograph.
IPTC is the descriptive, administrative and rights vocabulary: what the picture shows, who made it, who owns it, how it may be used. It originated in news-agency practice and is maintained by the International Press Telecommunications Council.
XMP is the container. Modern software stores IPTC properties as XMP inside the file, using identifiers such as dc:creator or photoshop:Credit. An older binary container, IIM, is still read by some software as a fallback, which is why good tools write both for the handful of fields that exist in each.
Which IPTC fields should an archive fill first?
If a project can only afford five fields on every file, these are the five. They are also the ones written by the free Add Copyright Metadata tool, which edits a JPEG's XMP without recompressing the picture.
| Field | XMP identifier | What it answers |
|---|---|---|
| Creator | dc:creator |
Who made the photograph |
| Copyright Notice | dc:rights |
Who claims copyright, in the usual © form |
| Credit Line | photoshop:Credit |
How the picture should be credited when published |
| Web Statement of Rights | xmpRights:WebStatement |
A URL where the rights position is set out |
| Rights Usage Terms | xmpRights:UsageTerms |
A plain-language statement of how it may be used |
Creator and Copyright Notice are different facts and often different names. In a single-photographer archive the creator never changes, while the copyright owner may be a company. Credit Line is different again: it is the wording a publisher should print, and it is the field picture desks look for.
The descriptive fields: caption, keywords, date and place
The second group is what makes a picture findable.
- Description (
dc:description): the caption. Who, what, where, when, in sentences. A separate guide covers how to write captions from contact sheets and sleeve notes without claiming more than the evidence supports. - Headline (
photoshop:Headline) and Title (dc:title): a short synopsis and a short reference name respectively. Title is not the caption. - Keywords (
dc:subject): the search terms. IPTC revised the definition of this property in version 2024.1 (December 2024). What goes in it is the subject of the companion piece on keywording with controlled vocabularies. - Date Created (
photoshop:DateCreated): when the photograph was taken, not when it was scanned. The property accepts a year or a year and month alone, which matters for film where the day is unknown. - Location Shown in the Image and Location Created (
Iptc4xmpExt:LocationShown,Iptc4xmpExt:LocationCreated): the place depicted and the place the camera stood. For a studio still life only the second applies. - Person Shown in the Image (
Iptc4xmpExt:PersonInImage): names of identifiable people, kept out of the keyword field so that a search for a person is exact.
Fields that matter most for scans of film
A film archive has a problem a born-digital library does not: every file is a reproduction of a physical object, and the object has its own identity. Three parts of the standard deal with this.
Digital Source Type (Iptc4xmpExt:DigitalSourceType) says how the digital image came to exist. The IPTC vocabulary includes negativeFilm (digitised from a transparent negative), positiveFilm (digitised from a transparent positive) and print (digitised from a non-transparent medium), alongside digitalCapture for camera originals. The same vocabulary carries the values used to label generative AI output, which is one reason the field is now read more widely than it used to be. An archive of medium-format transparencies, negatives and contact sheets should set it on every scan.
Artwork or Object in the Image is a structure for describing the physical original. Its sub-fields include Source Inventory Number (Iptc4xmpExt:AOSourceInvNo), Physical Description (Iptc4xmpExt:AOPhysicalDescription), Date Created and Circa Date Created (Iptc4xmpExt:AOCircaDateCreated). This is where "6×7 colour transparency, box 14, sleeve 3" belongs, so that the scan can always be traced back to the piece of film.
Identifiers. Digital Image GUID (Iptc4xmpExt:DigImageGUID) and Image Supplier Image ID (plus:ImageSupplierImageID) hold the archive's own reference for the file. A stable identifier is the single most useful thing a licensing desk can find in a stray file, because it resolves every other question with one database lookup.
What Google Images reads, and what earns the Licensable badge
Google's Search Central documentation lists six IPTC fields it recommends: Copyright Notice, Creator, Credit Line, Digital Source Type, Licensor URL and Web Statement of Rights. Creator, credit and copyright information can be shown alongside the image in Google Images.
For the Licensable badge the rule is narrower. Google accepts either structured data on the page or IPTC metadata in the file, and says only one is needed. By the IPTC route, Web Statement of Rights is the field that must be present. By the structured-data route, it is the license property on an ImageObject. Licensor URL (a sub-field of the PLUS Licensor structure) or acquireLicensePage adds the link to where the image can actually be licensed. If both methods are used and disagree, Google uses the structured data.
Eligibility is not a guarantee. The page still has to be crawled and the image indexed, and the badge is Google's to show or withhold.
Rights and AI fields worth knowing about
Version 2023.1 of the standard (October 2023) added Data Mining (plus:DataMining) and Other Constraints (plus:OtherConstraints), which let a rights holder state a position on text and data mining, including AI training. What that signal can and cannot do is covered in the article on signalling AI data-mining preferences in image metadata.
Version 2025.1 added four properties for images made with generative systems: AI System Used, AI System Version Used, AI Prompt Information and AI Prompt Writer Name. An archive of photographs should leave these empty. Their absence, with a film Digital Source Type present, is itself informative.
The standard also contains Model Release Status and Property Release Status. Fill them only from documents you hold. "None" and "Not Applicable" are valid values, and an honest "None" is more useful to a buyer than a blank.
A minimum record, and its limits
For a first pass over a large collection, a defensible minimum is: Creator, Copyright Notice, Credit Line, Web Statement of Rights, Description, Date Created (to whatever precision is known), Keywords, Digital Source Type and an identifier. Everything else can follow as cataloguing deepens.
Two cautions. First, embedded metadata is fragile. Many platforms and image pipelines remove it on upload or resize, and tools that re-encode an image through the browser canvas discard it too. The reverse operation is sometimes right: before sharing a personal photograph you may want to remove EXIF and GPS data while keeping the copyright notice. Either way, the catalogue database must hold the authoritative record, with the file carrying a copy.
Second, metadata is a claim, not a proof. Anyone can type a name into a Creator field. Its value comes from consistency across a collection and from agreement with the paper record.
Where this meets the Robert Harper archive
The Robert Harper Photograph Library holds 95,487 unique photographs by the archive's inventory count (a figure not yet checked frame by frame against the physical material), of which 52,684 are analogue and largely await scanning and cataloguing. That puts the archive at the point where field choices are still open, which is the cheapest moment to make them well. Picture libraries and distributors who want to discuss delivery specifications and metadata mapping can start on the page for picture agencies, and all of the browser tools mentioned here are collected under free image tools.
Frequently asked questions
What are the most important IPTC fields for a photo archive?
Start with Creator, Copyright Notice, Credit Line, Web Statement of Rights and Description. Those five tell anyone who finds the file who made it, who owns it, how to credit it, where to license it and what it shows. Then add Date Created, Keywords, Location Shown, an identifier such as Image Supplier Image ID, and Digital Source Type.
Which IPTC fields does Google Images use?
Google's documentation lists Creator, Credit Line, Copyright Notice, Web Statement of Rights, Licensor URL and Digital Source Type. Creator, credit and copyright can be shown beside the image. Web Statement of Rights is the field that makes an image eligible for the Licensable badge when you rely on IPTC rather than structured data. Where the two conflict, Google uses the structured data.
What is the difference between EXIF, IPTC and XMP?
Exif is mostly written by the camera or scanner: exposure, device, capture time. IPTC is the descriptive and rights vocabulary: caption, creator, copyright, keywords. XMP is the container most modern software uses to store IPTC properties inside the file. An older IPTC container, IIM, still exists and some software reads it as a fallback.
How should a scan of a negative or transparency be labelled in IPTC?
Set Digital Source Type to negativeFilm (digitised from a transparent negative) or positiveFilm (digitised from a transparent positive), and use print for a scan of a print or Polaroid. Record the original's creation date in Date Created, not the scan date, and use the Artwork or Object fields for the physical item's description and inventory number.
Does embedded IPTC metadata survive uploading to websites?
Often not. Many social platforms, content delivery networks and image-processing tools remove some or all embedded metadata when they resize or re-encode a file. Treat embedded fields as a label that travels with the file when conditions allow, and keep the full record in your own catalogue database.
Sources
- IPTC Photo Metadata Standard (overview and version history)
- IPTC Photo Metadata Standard 2025.1 specification
- Google Search Central: Image metadata in Google Images
- IPTC: Quick guide to IPTC Photo Metadata and Google Images
- IPTC NewsCodes: Digital Source Type vocabulary
Published by Archive.London. Corrections are welcome under our editorial policy.

