Metadata and provenance

Provenance for photographs: what buyers ask for

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.

A photograph has three histories: the object's, the rights' and the file's. Different buyers care about different ones, and an archive that can document all three, gaps included, is far easier to deal with.

Ask a print dealer, a picture editor and a dataset buyer what "provenance" means and you will get three different answers, all correct. The dealer means the history of an object. The editor means who took the picture and who can license it. The dataset buyer means where the files came from and on whose authority. A photograph has all three histories, and an archive should be able to give an account of each.

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

What does provenance mean for a photograph?

The object. Where the negative, transparency or print has been since it was made: the photographer's studio, a laboratory, an agency's files, a later owner. For prints sold as objects, this is the classic meaning, and it includes the difference between a print made close to the date of the negative and one made later, along with stamps, signatures and inscriptions.

The rights. Who was the author, who owns the copyright now, and by what route. The physical object and the copyright are separate property. Possession of a box of negatives does not by itself show ownership of the copyright in the images, which is why careful buyers ask for evidence of each.

The file. How the digital image was produced: from which original, by whom, when, and what has been done to it since. This history is new, and for film photographs it begins at the scanner.

What do picture agencies and publishers ask?

Their concern is whether they can publish safely and credit correctly. The usual questions:

  1. Who is the photographer, and is authorship consistent across the collection?
  2. Who owns the copyright, and does the person offering the licence have authority to grant it?
  3. Who and what is shown? Are there identifiable people, trade marks, artworks, designs or private property in the frame, and what is known about any permissions?
  4. Where has the image been published before, and are there earlier licences or exclusivities still running?
  5. What credit line should be used?
  6. Is the caption reliable, and what is it based on?

The third question is the one a single-owner archive must answer with most care. One rights holder for the photographs means one counterparty for the copyright in the images. It says nothing about the rights of people or brands depicted, which have to be considered use by use. The method behind a reliable answer to the sixth is covered in writing captions from contact sheets and sleeve notes.

What do curators and collectors ask?

Institutions look at the object and its documentation.

  • Is this the original camera film, a duplicate, or a later print?
  • What annotations, stamps or labels does it carry, and in whose hand?
  • What condition is it in, and how has it been stored?
  • Has it been exhibited or published, and where?
  • Is there an inventory, and does the item's number tie it to a described group?
  • Is the surrounding material, such as contact sheets, job records and published pages, kept with it?

That last point is often decisive. An archive kept together, with its working papers, can explain itself in a way that dispersed prints cannot.

What do AI and data buyers ask?

Dataset licensing has added a new set of questions, driven partly by regulation. Under Article 53 of the EU AI Act, providers of general-purpose AI models have, since 2 August 2025, been required to put in place a policy to comply with EU copyright law and to publish a sufficiently detailed summary of the content used for training, using a template issued by the AI Office. The Commission's enforcement powers have applied since 2 August 2026. The summary is a description on a template, not a work-by-work list, but a provider who must describe its sources has reason to prefer sources that can be described.

In practice a data buyer's diligence tends to cover:

Question Why it is asked
Who created the images? A single known author is simpler to document than an aggregated pool
Who holds the rights and can sign? The licence is only as good as the licensor's authority
How were the files obtained? Direct scans from originals are traceable; scraped copies are not
Have the images been widely published online? Affects novelty of the data and overlap with existing training sets
Are identifiable people depicted? Raises privacy and likeness questions separate from copyright
Are third-party brands, artworks or designs shown? May limit some uses
What metadata comes with each file? Captions, dates and categories determine what the data is useful for
Is there a rights reservation on the files? Shows the owner's stated position and where to ask

Nothing in that list is satisfied by a label. No responsible archive should describe its images as pre-approved for training in the abstract. The accurate statement is that licences are negotiated directly with the rights holder, subject to rights review, with scope set by what the review finds.

What evidence answers these questions?

Evidence What it supports Where it lives
Original film with sleeves, mounts and annotations Authorship, date, sequence, authenticity The physical archive
Contact sheets and job records Occasion, selection, client The physical archive
Published pages with credits Use, date, attribution The physical archive; libraries
Inventory and collection descriptions Extent, arrangement, identifiers Catalogue database
Written evidence of copyright ownership Authority to license Owner's records
Scan log: operator, date, equipment, source identifier Origin of each file Digitisation records
SHA-256 hash of each master file Exact identity of a file over time Hash register kept outside the file
Embedded IPTC rights and source fields Claims that travel with the file Inside the file
Content Credentials manifest Who signed the file and whether it has changed Inside or alongside the file

The top half of the table is analogue and cannot be manufactured later. The bottom half is cheap to create at the moment of scanning and expensive to reconstruct afterwards.

What can a digital file carry, and what can it not?

Three mechanisms are worth using, each with a limit.

IPTC metadata. Creator, Copyright Notice, Credit Line, Web Statement of Rights, Digital Source Type and the Artwork or Object fields can state who made the photograph, who owns it and which piece of film the scan came from. The EXIF and metadata viewer shows what a file contains, and the Add Copyright Metadata tool writes the main rights fields to a JPEG without recompressing it. The limit: anyone can type anything, and many platforms strip it.

Content Credentials. A signed C2PA manifest shows which certificate holder vouched for a file and detects later alteration. The Content Credentials checker will show whether a file has one. The limit, discussed in Content Credentials explained for photo archives: a credential starts at signing and cannot attest to events before the file existed.

Hashes. A SHA-256 value computed when a master scan is made, for instance with the image hash and fingerprint tool, lets anyone later test whether a file is bit-for-bit that master. A perceptual hash helps find resized or re-encoded copies. The limit: a hash identifies a file, not its author.

A fourth signal, the IPTC Data Mining property, states the owner's position on AI training inside the file. It is described in signalling AI data-mining preferences in image metadata.

Saying what is not known

The most persuasive provenance statements include their gaps. A count that is described as coming from an inventory and not yet checked frame by frame against the physical material is more credible than a round number with no source. A caption that says "circa" is more useful than a confident wrong year. Buyers of every kind are used to incomplete records. What they cannot work with is uncertainty that has been hidden.

How the Robert Harper archive answers

The Robert Harper Photograph Library is the work of one photographer across approximately thirty-five years, from two years assisting David Bailey, Helmut Newton and Lester Bookbinder at the Condé Nast studios to his own studio and commercial clients that included De Beers; the catalogue includes a collection titled De Beers diamonds. The inventory records 95,487 unique photographs, 52,684 of them analogue and 42,803 digital, or 177,143 frames including duplicates and variants. Those counts are the archive's own and have not yet been independently reconciled, and collections may overlap. The archive is owned by 825 Ventures Ltd, the owner and rights holder.

The figures are set out on the archive facts page and the approach to rights under rights and stewardship. The analogue material largely awaits digitisation, so the file-level record described above is still to be built. Agencies, curators and data teams with specific diligence questions are welcome to put them through professional access.

Frequently asked questions

What does provenance mean for a photograph?

It is the documented history of the photograph: who made it and when, where the physical original has been held, who owns the copyright and by what route, and how any digital file was produced from it. For prints sold as objects the emphasis is on custody and authenticity; for licensing it is on authorship and rights.

What documents prove the provenance of a photo archive?

Typically an inventory, the original negatives or transparencies with their sleeves and annotations, contact sheets, job records and correspondence, published pages showing credits, and written evidence of who owns the copyright. For digital files, add scan logs, stable identifiers and file hashes. No single document is sufficient; the strength lies in their agreement.

Not automatically. The physical object and the copyright in the image are separate property, and they can be held by different people. A buyer licensing reproduction rights will want to see evidence of copyright ownership specifically, not only possession of the film. This is general information, not legal advice.

What provenance do AI companies ask for when licensing images?

Commonly: who created the images, who holds the rights and has authority to license them, how the files were obtained, whether identifiable people or third-party brands appear, and what documentation accompanies the dataset. EU rules requiring model providers to publish summaries of their training content have made a traceable source more valuable than it was.

Can metadata or Content Credentials prove provenance?

They can carry provenance statements and, in the case of Content Credentials, show who signed a file and that it has not changed since. They cannot prove that the statements are true or reach back before the file existed. For film photographs, the physical original and the paper record remain the primary evidence.

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