We labelled sites “urban infill” with our own rule. For most of them (201 of 232) it is a land-cover test (outside the Green Belt, low tree canopy), not anything the campaign said; for 31 it is a keyword in the campaign name or tags. That label implies reuse of previously developed land. The government’s own register disagrees for 93.1% of the sites that carry it, but the label itself has not yet been checked against a second coder.

What we measured

We checked each site against the national Brownfield Land Register, counting a site as registered if it falls on or immediately beside a registered parcel. Every coordinate was validated against an England bounding box first, because the platform’s own documentation records swapped coordinates that place sites in the sea or in Ireland.

What it shows

31 of 652 sites (4.8%) are on the register. For urban infill and estate redevelopment it is 16 of 232 (6.9%), against 15 of 420 (3.6%) for everything else.

That difference is borderline: Fisher exact p = 0.082 and chi-square p = 0.056. It is a comparison chosen after looking at the data, so treat it with caution. Across all four development types together the test is p = 0.14. Whichever way it is read, registered brownfield reuse is the exception everywhere, infill included.

Read it carefully

The register is published by councils, and the team that builds it has said in writing that council data was inconsistent and hard to aggregate despite a 2017 duty to publish. So “not on the register” is a fact about the register. It doesn’t prove the land was never built on. What it does remove is the register-backed claim to “brownfield first” that the infill label implies.

Revised 12 November 2026: The difference between infill and other sites is now described as borderline (Fisher p = 0.082), not as no difference. Added whose label “infill” is.

Corrected again 12 November 2026 after a check of the data pipeline: the description of how “infill” is assigned is corrected and marked as untested.