Search beyond the first few words
Keyword search now examines the full indexed text of each record rather than a short initial token list. A dedicated search index covers reviewed retained snapshots and current dataset views, with all 178 reviewed datasets represented in discovery. The initial release contains over 900,000 indexed records and field groups.
The search index refreshes on an hourly schedule. Source observation and snapshot dates remain distinct from the index refresh time, so rebuilding the index does not make old evidence new.
More useful keyword matching
Search supports common abbreviations such as US, GDP and BTC, plural forms, accented names, partial words and conservative spelling corrections. Quoted phrases provide exact phrase matching. Titles receive greater relevance weight, and canonical duplicate results are grouped.
Matching terms are highlighted safely in results. Search explains spelling and prefix matches, retains category filters and pagination, and lets eligible users open a result's workspace directly from Simple mode.
Coverage you can see
Results show dataset coverage, including catalogue-only and bounded sources. At initial verification, 177 of the 178 reviewed datasets had indexed snapshots; the remaining infrastructure-feature endpoint requires an individual feature ID. Retained source views and source pagination can be partial. Raw map geometry and embedded binary imagery are not keyword-indexed.
Your account's source entitlements continue to control results, counts and spelling suggestions. Search uses shared reviewed evidence; private account, payment and personal-workspace data are excluded.