The summary, the facts behind it, and how many people read it

The QID. A measured summary of Ada Lovelace returned title "Ada Lovelace", pageid 974 and wikibase_item "Q7259" side by side. Titles get renamed by editors and differ per language. pageids survive renames but are scoped to one language edition, so 974 means something completely different on the German wiki. Q7259 is the same human on every one of the 360+ editions, which is what makes it the only safe join key — and langlinks turns it back into whatever title you need.

Because the two shapes serve different code. key_facts is a normalized set with predictable snake_case keys — instance_of, born, died, birthplace, deathplace, citizenship, occupation, spouse, father, mother, children — which is what you bind to a UI. facts is the raw Wikidata property set with human labels and spaces in the keys: "birth name", "place of birth", "field of work", "Commons category". Measured, both were present in one response. Use key_facts unless you need a property the normalized set does not cover, then fall back to facts.

Sometimes. In a single measured response, sex_or_gender was the string "female" while occupation was an array of eight entries and children was an array of three. The type follows the underlying Wikidata property's cardinality, so a person with one occupation and a person with eight do not return the same shape. Normalize every value to an array on the way in, or your renderer will print a string character by character the first time it meets a single-valued field. value_cap limits how many entries a multi-valued property returns.

They are Wikidata statements passed through unchanged, which means they are as good as Wikidata is — and Wikidata contains contested, outdated and occasionally plainly wrong claims, especially for historical figures where statements get attached loosely. We do not correct or filter them, because a silent correction would leave you unable to tell our guess from the source. If a fact matters, the QID gives you the Wikidata item to check the statement's own references.

Two, in the same response, so read carefully. A measured 7-day pull returned start "20260819" and end "20260826" — compact YYYYMMDD with no separators — while every entry inside daily[] used ISO "2026-08-19". Alongside the series you get total 26637 and average 3805.3 precomputed. The response also echoes project, granularity, access and agent, which matters because the default agent is all-agents: that figure includes bot traffic, and filtering to human users gives a materially lower and usually more honest number.

It is a placeholder for a location Wikipedia could not attribute to a country, not a real ISO code. A measured search around Istanbul returned three results within 41 metres: a landmark with country "TR", an event page with country "XX", and a third with country null. Filter on it and you will silently drop historical events, undersea features and disputed places. dist_m is a float in metres and is reliable — a result at dist_m 0 sits exactly on the coordinate you passed.

Because the subject is not a place. Measured, the summary and the entity record for a person both returned coordinates null, while geosearch results carry lat and lon directly. Only articles about a physical location have coordinates, so treat null as the normal case and use type to tell what kind of article you are looking at — the measured summary reported type "standard", and disambiguation pages and redirects report their own values.

Pass lang on any action for a direct read, and use langlinks when you want to know which editions exist for a subject before you ask. The QID from a summary is the reliable bridge: get Q7259 once, then resolve it wherever you need it rather than translating the title yourself. Coverage depth varies enormously between editions — an article that is comprehensive in English can be two sentences elsewhere — so check the extract length before you treat a translation as equivalent content.

Wikipedia API is a ReefAPI endpoint group for article summaries, search and structured facts. It returns live JSON through POST requests under /wikipedia/v1.

Yes. ReefAPI starts with 1,000 free credits, no card required. Wikipedia calls use the same shared credit balance as every other ReefAPI engine.

No login to Wikipedia is needed for the API response. You call ReefAPI with your x-api-key header, and the playground can run live examples before you create a production key.

The page example is captured from a live summary call, and production requests fetch live data through ReefAPI rather than a static sample.

Wikipedia actions currently cost 1-2 credits per successful call. Failed or blocked calls are free, and all APIs draw from one credit pool.

Yes. Connect ReefAPI once through MCP and your assistant can call wikipedia actions with the same key, credit pool and JSON envelope used by normal REST requests.