How we source our data

We'd rather explain this plainly than let you guess. Everything on a care home's page traces back to one of three sources: the CQC's own register, the CQC's own inspection reports, and Ordnance Survey-derived postcode data. We add a layer of automated summarisation on top of the inspection reports, described below, but we don't collect or estimate anything about a home that the CQC and the Post Office haven't already published.

Registration and ratings

Names, addresses, registration status, bed capacity, care types, specialisms and CQC ratings (overall and by category) come from the Care Quality Commission's Syndication API and bulk data exports, published under the Open Government Licence. We pull this data on a schedule, not once — a "data last verified" date on every listing page reflects when we last checked it against the CQC's own records, not when the page happened to be built.

Location

Coordinates are derived from each home's registered postcode via postcodes.io, a free, open geocoding service built on Ordnance Survey and ONS data. We use this to place a home on the location diagram on its own page and to work out genuinely nearby homes of a similar type — not to estimate anything about the property itself.

Inspection summaries

A CQC inspection report typically runs to several thousand words of regulatory language. For homes where we've completed this step, we generate a plain-language summary of what the report actually found, using a large language model working directly from the official report text — not from the rating alone, and not from any other source. The summarisation prompt is explicitly barred from stating a rating word (Outstanding, Good, Requires improvement, Inadequate) inside the summary itself, so the written text can never drift from what the rating badge on the same page shows; every generated summary is automatically checked against that rule before publication, and anything that fails is regenerated. Where we haven't yet produced a summary, the page says so — we don't fill the gap with invented text.

This is a machine-assisted process today, not yet a fully human-reviewed one for every home. We consider that an honest limitation to state rather than paper over, and it's why we treat the "report an error" link on every home's page as a real feedback channel, not a formality — see our editorial policy for how corrections are handled, and how we plan to bring a named care-sector reviewer into this process as the directory grows.

What we deliberately don't collect or estimate

• We don't publish photographs we haven't been given by the home's provider — no stock imagery, no AI-generated images standing in for a real building.
• We don't estimate a specific home's fees. Where we show a cost range, it's a regional estimate, clearly labelled as such, never a claim about what any particular home charges.
• We don't collect anything about individual residents. Nothing on iCare concerns identifiable people receiving care.

Attribution

Data sourced from the Care Quality Commission, licensed under the Open Government Licence v3.0. iCare is not affiliated with or endorsed by the CQC.