Of roughly 5,600 music-related websites in the National Library of Scotland's Scotland on the Internet archive, nearly half no longer existed on the live web when researchers built an exhibition from the collection.

The number is startling, but it is easy to misread. It does not mean half of Scotland's music web has disappeared. The archive is not a census of everything that was ever online. It means that among the sites the archive did collect, a very large share had since vanished from the web people could ordinarily reach.

Digital Ghosts: Visualising Scotland's Web Archives won the Digital Preservation Coalition's 2026 award for teaching and communications on October 15. The project turns web-archive metadata into interactive visualizations and physical artworks. Its subject is web decay, but the installations keep returning to a harder problem: what exactly can an archive say about material that is absent from it?

A missing website is not one fact

Suppose an archived page from 2012 now returns an error on the live web. The archive can establish something useful. The page existed, it was captured, and the live version later disappeared.

Now suppose a search of the archive returns nothing.

That blank is much harder to interpret. Perhaps the site never existed. Perhaps it existed but was never selected for capture. The crawler may have failed to reach it. The page may have changed between captures. A preserved copy may even exist with metadata too poor for the query to retrieve it.

Those cases produce the same immediate experience for a researcher: nothing appears.

One Digital Ghosts installation, Sites: Surfaced or Submerged, shows when sites entered the collection and whether they remain available on the live web. It also makes changes in collecting attention visible. Religion is prominent in some periods; sports, arts, and lifestyle material become more prominent in others.

Curated archives necessarily contain decisions like these. Selection policies, staff time, technical limits, and circumstance all shape the record. The collection consequently contains evidence about the web and about the institution that collected it.

A second installation, Pandemic in the Archives!, uses more than 1,000 pandemic-related sites and news items from 2019 to 2025. Here the difficulty is temporal. A page can stay at the same address while its content changes repeatedly. If the crawler visits in March and June, material that appeared only in April may never become part of the archive.

For historians of the web, knowing that a URL was preserved is only the beginning.

Search boxes make collections look cleaner than they are

A normal archive interface has a sensible job: find the item requested. In doing that job, it tends to hide the machinery of collection. Search results do not usually show failed crawls, uncaptured revisions, shifting curatorial priorities, or the unknown territory outside the collection.

Andrea Kocsis and Dorsey Kaufmann's 2026 paper on Digital Ghosts treats messy metadata as material worth examining. Gaps and inconsistencies can carry information about the web's instability and about institutional choices over what becomes cultural memory.

Visualization helps because some of those facts exist at collection scale. A list of URLs can answer whether a particular site is present. A map or time series can reveal that one region is densely represented and another thinly, that one class of sites disappears faster than another, or that a collecting priority arrives suddenly. Those patterns are difficult to notice one record at a time.

The award-winning project also includes a work that abandons the screen altogether. Over a 120-hour performance, ultraviolet projections tied to website lifespans and topics developed cyanotype prints on a gallery wall. When the light stopped, the wall retained uneven blue traces.

A search result normally suppresses that kind of unevenness. Here it was left where visitors could see it.

Sources

Ulix app

Keep Clip

Keep Clip is an Android capture tool for saving text, quotations, highlights, and links, with search and export to ordinary file formats.