Home · Wikimedia
Wikimedia Commons · Research strand

Open heritage on Wikimedia Commons

What national libraries open, where it travels, and what generative AI does to its value — a three-paper programme on the public value of digitised public-domain heritage.

Three national libraries have each uploaded tens of thousands of public-domain images to Wikimedia Commons. We built two large, original datasets on the same images to ask three questions in turn: what gets opened, where it travels, and what generative AI does to its value.

0
files in the Wikimedia census
0
detected reuses — from a random sample of 15,000 images
0
national libraries — UK · NL · CH

What we study

The same heritage image can be given away for free on Wikimedia, sold as a stock photo for a licence fee, and reused on social media — all at once. We follow each uploaded image across the open web to see where it goes and what it is worth.

The same public-domain image shown three ways: free on Wikimedia Commons, sold for a licence fee on a stock-photo site, and reused on social media.
The same image, three worlds — free on Wikimedia Commons (left), sold on a commercial stock site (centre, ~$193), and reused on social media (right).
Paper 1 · the provision sideInvited · special issue

Open by Design — putting it online is itself a strategy

Research question: How do national libraries enact a digitisation strategy through what they actually upload — and what does that mean for public value?

From the outside, uploading public-domain images looks like one good deed. Look at what each library actually uploaded — the format, the rights labels, the metadata, the timing — and three very different strategies appear.

British Library

Decentralised mass-release — huge volume, ambiguous rights, 52 % templated metadata.

Koninklijke Bibliotheek

Curator-led — steady, clean public domain, its oldest holdings.

Swiss National Library

Centralised uploads; metadata straight from the archive database — mostly TIFF-format access derivatives (processed and downscaled for the platform, not archival masters), multilingual, 95 % attributed.

The lesson: format, licence and metadata are not paperwork — they decide how much public value each upload actually realises.
Paper 2 · the diffusion sideUnder review · Journal of Cultural Economics

From Digitization to Diffusion — where open heritage travels

Research question: Once heritage is on Commons, where does it actually go — and which upload choices make an image diffuse?

We followed 15,000 library images out into the wild and found them again 173,000 times — on stock sites, in news stories, on Wikipedia, on social media, across dozens of countries. Roughly two in three images were reused at least once.

Where reuse happens — by library
detected reuses by country · hover a country for the exact count

    Colour intensity = detected reuses in that country. Reuse is strongly European-centred; the Swiss National Library peaks at home, the KB in the Netherlands, while the British Library reaches furthest into English-speaking countries. Excludes an aggregated “European Union” label.

    0
    of images reused at least once
    0
    detected reuses
    $181–423m
    modelled avoided licence cost

    What drives reuse (image-level analysis)

    Public domainStrongest positive lever — commercial & news reuse
    Strong ▲
    JPEG access formatReused far more than TIFF files
    Years onlineEach year online ≈ +29% odds of reuse
    +29% / yr
    Very high resolutionLess likely to be reused at all…
    …once adoptedHigh-res images are reused more intensively
    ◆ intensity

    Direction from the image-level analysis; bar widths indicate relative strength, not exact odds. Only “years online” is quantified precisely.

    Open images travel far — but design decides how far. A clear licence and a friendly file format spread heritage the most.
    Interactive · built on the paper's own valuation

    What is this open heritage worth?

    Most detected reuses are commercial — the image reappears on a stock-photo or resale site that shows a price. Those are the reuses that would otherwise have paid a licence fee. Pick a reuse channel to see how each behaves, then value the commercial reuses at a Getty-style licence price.

    $150
    $0$150 · low-res Getty$350 · high-res
    $181m
    estimated avoided licence cost · ≈1.21m commercial reuses × $150
    British Library$129m
    Koninklijke Bibliotheek$37m
    Swiss National Library$16m
    Select “Commercial / stock” above to value these reuses.

    The paper benchmarks commercial reuse against a standard Getty Images licence — USD 150 (low-resolution) to 350 (high-resolution). At those prices the three collections represent a modelled $181m–$423m in avoided licensing cost — an explicit lower bound, since it excludes all non-commercial, educational and AI-training value.

    Paper 3 · heritage in the AI economy

    Stock or Prompt

    Since late 2022, anyone can generate a convincing picture from a sentence. So we ask a simple question of our heritage images: when synthetic images become free, does the world still want the real, authenticated one? Early evidence suggests demand splits — falling where any old-looking image will do, rising where it matters that the picture is genuine.

    Work in progress · full results coming soon

    See all publications →

    Enlarged figure