Blog · 18 August 2026
Published is not the same as findable
UK public bodies already publish minutes, spend and registers. The gap is access: image PDFs, split portals, and joins that nobody can make by hand.
Public bodies in the UK already publish. Committee minutes go on a ModernGov site. Spend files sit on a transparency page. Registers of interests appear as PDFs. Companies House, the title register and the charity extract are all official. The gap is not a missing law. The gap is that published is not the same as findable.
A scanned image that cannot be searched is not access. A CSV three clicks inside a finance page is not access for most people. A name that appears in a minute, a filing, a title and an island paper is visible only to someone who has looked in four places and joined them by hand. Most people cannot do that. Most newsrooms cannot do it at scale.
What fragmentation looks like
Take an invented council, Vale Civic. Its cabinet minutes are image PDFs. Its officer list is a Companies House filing under a local company. A hospitality lease sits on the England and Wales title register against an overseas proprietor. A related paper exists on a Crown Dependency site. None of those publishers is hiding the file. None of them will answer a single query across the set.
Official portals are built for the body that owns them. They are not built for the reader who needs the join. That is a structural problem, not a scandal about Vale Civic.
What Institrace does with that
Institrace collects the papers that are already public, stores a hash at the moment of indexing, and runs OCR on image PDFs so the text exists as text. Officers, land titles, contract awards, charities and NHS organisation codes sit on the same graph. You can search it. You can ask a question of it. Every answer is meant to point at the original file.
We do not add facts. We do not allege. We do not treat a structural flag as a finding. The index makes the published record usable. What you do with it stays with you.
What you can see without a login
The homepage counts, the Sources catalogue, a sample of recent titles on each source page, and the product pages: Capabilities, About, this blog, the FAQ and Resources. Full-text search and the People, Money, Land and Flags indexes need a signed-in seat.
If a source page shows fifteen titles and a count of thousands, the count is real. The sample is what search engines and logged-out readers get. That is deliberate. The catalogue should be crawlable. The corpus should not be scraped as a dump.
Read next
For the NHS layer, start with what we hold on the NHS and why an ODS code is the join. For a short answer to a specific question, use the FAQ.