About the data
Where these 5,000 years of events come from, how each one was checked, and what we still cannot promise.
What is in here
5,146 events, each with a time, a place, a short summary in three languages, and a list of sources. Coverage runs from the Neolithic to the year 2000, across ten world regions.
Events are assembled by an automated pipeline, not typed in by hand. Every record passes format checks, a distribution report, an AI review pass, and a frozen regression sample before it is published.
Sources and licences
Summaries are written from verified facts, not copied. The underlying evidence comes from:
- Wikidata CC0 1.0structured facts — dates, coordinates, classifications
- Wikipedia CC BY-SA 4.0narrative text used as evidence for dates and places
- UNESCO World Heritage Centre © UNESCOofficial World Heritage listings — the only third-party source independent of the Wikimedia family
- OpenStreetMap ODbLbase map tiles
Wikipedia text is CC BY-SA 4.0 and OpenStreetMap data is ODbL; both require attribution, which this page provides. Our own summaries are original prose written from checked facts.
How a record is checked
Dates and places are normalised first: a date known only to a century is stored as a century, never as a false exact year. A location that is schematic rather than a precise site is marked as such on the map.
Records that fail a structural rule are held back rather than published with a guess. Where sources disagree beyond reconciliation, the record is withheld — we would rather have a gap than a wrong date.
What we cannot promise
454（9%） events have at least one source independent of the Wikimedia family. The remaining 4,692（91%） rest on Wikidata and the Wikipedias, which count as a single source family.
So this is not a cross-verified dataset. It is a sourced one: every fact tells you where it came from, and you can check it yourself.
Found something wrong? That is useful — the pipeline is rerun regularly and corrections feed back into it.