FAQ
Frequently asked questions
If something here is not covered, the guide walks through the API in workflow order, and support answers the rest.
Is GeoAPI actually free?
Yes, entirely. There is no paid tier, no subscription, and no dataset held back behind a payment. Registration exists solely to issue an API key.
What exactly do I need an account for?
Only for the endpoints that read your own account — usage figures and key management. Every dataset endpoint (areas, population, coverage, places, tiles) and the public map work with no account and no key at all.
What do you log about my requests, and for how long?
For authenticated requests: the endpoint, method, status, response time, your IP address, your user agent, and which key and account made the call. Request bodies and responses are never stored. Those rows are deleted automatically after 30 days — the log exists to debug failures and detect abuse, and neither needs a longer memory than that. Keyless calls to the public dataset endpoints are rate-limited per address but are not written to that log at all.
Do I need an API key to use the map?
No. The map and every informational page are public: no sign-in, no key, no session of any kind.
Where do the population figures come from?
From published third-party datasets, ingested with provenance intact: administrative boundaries from one class of provider, modelled population from another, official projections from a third. Every provider currently loaded, what it contributes, its licence, and its required attribution text are listed on /sources.
Why does one area have several different population figures?
Because the providers behind them measured different things. An enumerated census, an official projection carried forward from a census baseline, and a gridded model built from settlement imagery are three genuinely different measurements of the same territory, and they can differ substantially. Rather than hiding that, the API returns all of them and marks which one was selected as the headline figure.
How is the headline figure chosen?
By a fixed precedence rule applied identically everywhere: measurement method first (census outranks projected, which outranks modelled), then source priority, then the most recent reference year. It is never an average, and never simply whichever import ran last.
How accurate are the figures?
It depends on the method, which is why the method is returned on every observation. Enumerated census counts are the most reliable and the least frequently published. Modelled figures give even spatial coverage and hold up well in aggregate, but individual areas can diverge sharply from a reference count — coastal and island areas in particular. If accuracy for a specific area matters to your work, read the observation list and select on method rather than taking the resolved value.
Why do population or density sometimes show an em-dash?
Because no observation has been recorded for that area. The API omits those fields entirely rather than returning 0 or null: "not measured" and "measured as zero" are different claims, and collapsing them would quietly turn unsurveyed areas into empty ones.
Why is my country or region not covered yet?
The data model is country-agnostic by design, but coverage is built out region by region and is not uniform even within a covered country — top administrative levels are typically far more complete than the finest ones. Query /v1/coverage to see exactly which countries, levels, and sources hold data today rather than inferring it from an empty result.
Can I reuse the data? What are the licence terms?
Yes, including commercially, provided you reproduce each provider's attribution. The third-party datasets behind the population endpoints are permissively licensed but all require credit. Business listings served by /v1/places are first-party and proprietary: queryable through the API, but not for bulk extraction or resale as a standalone dataset. Exact attribution strings are on /sources — copy them verbatim.
If it is free, why are there rate limits?
To keep the service available to everyone. Limits exist for abuse prevention, not as an upsell — there is no paid tier to raise them into. Public dataset endpoints are limited per client address; keys carry their own per-key limit and monthly quota.
Is there an OpenAPI specification?
Yes, served at /openapi.yaml. Use it for client generation and schema validation; use the guide for the workflows a specification cannot express.