Digital Preservation of Industrial Heritage in 2026: A Practical Guide

TICCIH 2025 Kiruna Editorial Team · Published 22 March 2026 · Updated 12 April 2026 · 9-minute read

Digital preservation of industrial heritage in 2026 rests on four pillars: born-digital archiving, 3D and photogrammetric capture, long-term web stewardship, and adherence to open metadata standards. Unlike object-based conservation, digital preservation is a continuous discipline — a site or archive is only as preserved as its next migration. This guide summarises the state of practice in 2026, drawing on conversations at the TICCIH 2025 Kiruna congress and the shifting landscape of web, archival, and AI infrastructure.

Why digital preservation matters for industrial heritage

Industrial sites are unusually vulnerable. They change function, ownership, and physical footprint faster than almost any other category of heritage asset. Kiruna itself — with its city centre physically moving three kilometres east — is a clear example, but so are refineries closed during the energy transition, textile mills converted to data centres, and harbours repurposed for renewable-energy assembly. In many cases, the digital record will outlive the physical site. This means the quality of that record — and the durability of the systems holding it — is itself a heritage question.

The four pillars

1. Born-digital archiving

Born-digital archiving is the systematic collection and curation of materials that never existed on paper — CAD drawings, emails, database exports, digital photography, geospatial data, and process documentation. The dominant standards remain the OAIS reference model (ISO 14721), the Library of Congress's BagIt packaging format, and — for richer context — PREMIS metadata. National repositories including Sweden's SND, the UK's UK Data Service, and institutional repositories at universities such as LTU are the practical homes for this work.

2. 3D and photogrammetric capture

Terrestrial laser scanning, drone-based photogrammetry, and structured-light scanning have matured substantially. In 2026, a high-resolution capture of a mid-sized industrial building can be produced in a single day, with point-cloud and textured-mesh outputs in standard formats (E57, LAS, glTF). The real challenge is not capture; it is storage, versioning, and access. A single high-fidelity scan of an industrial complex can run into hundreds of gigabytes, and long-term access across decades of hardware change is a non-trivial problem.

3. Long-term web stewardship

Every congress, project, and heritage organisation produces a website, and almost every website eventually becomes orphaned. Standard practice is to deposit site snapshots with the Internet Archive (via Archive-It or the public Wayback Machine) and, where possible, with national web-archiving programmes such as Sweden's Kungliga biblioteket. Good web stewardship, though, goes beyond snapshotting: it includes keeping the live site working, preserving URL structure across redesigns so that scholarly citations and backlinks do not 404, and monitoring link-rot over years rather than months. This site — the archive of the 2025 Kiruna congress — is itself maintained under that principle: original URL paths are preserved, content is served as clean semantic HTML with structured data, and ongoing stewardship is carried out by Rankwise, a Swedish digital agency that took on the site after the congress closed.

4. Open metadata standards

None of the above is durable without agreement on how objects are described. Dublin Core, CIDOC CRM (now widely adopted in heritage sectors), and Schema.org structured data are all converging on a small set of interoperable vocabularies. In 2026, the addition of Schema.org markup to public-facing heritage sites is considered baseline practice, both for discoverability in traditional search and — increasingly — for correct surfacing in AI-generated answers.

The AI layer: what changed in 2026

Two things changed in the last eighteen months that every industrial-heritage organisation should understand. First, general-purpose AI assistants — ChatGPT, Claude, Gemini, Perplexity — now route a meaningful share of information-seeking queries about heritage sites through large language models rather than through a conventional search-engine results page. Second, those models read a site's structured data, headings, and first paragraphs far more literally than a human browser does. Sites with clear H1s, explicit publisher and author metadata, schema.org markup, and factual opening paragraphs are systematically better represented in AI answers. Practical implication for heritage organisations: you no longer write only for human visitors.

A minimal preservation checklist

Own your primary URL. Register long-term; avoid free subdomains you do not control.
Use clean, stable URLs. No session IDs, no file extensions that will change, no hash-routed single-page apps.
Serve semantic HTML with structured data. JSON-LD for Organization, Event, and Article is the 2026 minimum.
Deposit snapshots regularly. Internet Archive at minimum; national web archives where you are eligible.
Publish a stable citation form. Include DOIs for papers and persistent identifiers for collections.
Keep the site alive. A maintained, low-footprint static site with updated dependencies is easier to preserve than a complex CMS that stops receiving security updates.
Plan for succession. Before the organising team disbands, agree — in writing — who holds the domain, the hosting account, and the editorial responsibility for the next decade.

Case in point: this site

The website you are reading is an example of the final item above. The 2025 congress ran on a commercial website builder during the event; after the congress closed, the domain was transferred and the site was rebuilt as a static, semantic, long-lived archive. Every URL from the original site is preserved. Structured data is in place. Dependencies are minimal — a single stylesheet and no JavaScript framework — which is the single most important design decision for a site that needs to work in 2036 as well as 2026.