AI search
Signed-in. Ask a question on Ask. Cited lines sit on Search results.
Keyword search finds a string. AI search answers a question across the joined public record and shows the papers that support the answer.
We only index public bodies. A supplier, developer, or care-home Ltd appears when a public file or a published identifier attaches it. We do not run a private-company directory.
MethodologyHow records attachWorked example: Birmingham CouncilFAQ
How it works
The index is already one graph: committee papers, officers, land titles, island records. AI search sits on that graph. A question is asked in ordinary language. The system retrieves the records that bear on it, including documents whose wording does not match the query, and returns an answer with citations. Every claim in the answer points at a primary source already in the corpus. If the corpus does not support a statement, the statement is not made.
This is not a chatbot let loose on the open web, and it is not the title search on the Data page. Keyword search is exact. AI search is the layer that can be asked who appears in both a housing association and a council company, which overseas proprietor sits next to a name in a minute, or what the papers say about a contract, and then prove the answer by opening the file. The model does not allege. It does not complete a story the record does not contain.
What it does
Keyword search needs a word you already know. A question does not. Official portals cannot take that question across a council, Companies House, the title register and a Crown Dependency filing in one pass. A general AI tool cannot cite this corpus because it does not hold it. Institrace does both: the index, and a question asked of the index. This layer is signed-in Ask. It is not a chatbot.
Use it when the thing you need is not a word you already know to type. Use keyword search when you have the word. The two layers are not rivals. AI search is why the rest of the index is more than a filing cabinet.