Heritage has always been a spatial discipline. Every listing, every conservation area, every archaeological record is anchored to a place on a map. What is new is the volume of spatial data now available - and the arrival of tools that can read it at a scale no human team could match.
National LiDAR coverage, satellite imagery, crowdsourced photography and digitised historic maps together form an extraordinary record of the historic environment. Machine learning is beginning to make that record legible: identifying lost field systems, flagging unrecorded earthworks, and monitoring the condition of buildings at risk.
From record to insight
The opportunity for heritage organisations is not the technology itself but the questions it lets us ask. Where is heritage most at risk from climate change? Which communities live furthest from an accessible historic site? How has a designed landscape actually changed since its registration?
The most valuable output of a heritage AI project is rarely a map - it is a better decision.
These are planning questions, funding questions and audience questions. Answered well, they strengthen National Lottery Heritage Fund applications, sharpen interpretation, and target activity plans at the people a project most needs to reach.
Proceeding with care
There are real caveats. Models trained on incomplete records reproduce the gaps in those records; automated detection is a prompt for expert judgement, not a substitute for it; and communities must remain the authority on what their heritage means. The sector's task is to adopt these tools deliberately - with the same rigour it applies to any other survey method.
This piece sketches where the tools are today, what they realistically cost, and where a small heritage organisation should - and should not - invest.
