GET /doc/{id}

GET https://api.unzoi.com/doc/{id}

One article's full metadata.

Names, structured locations and dates, quotations, amounts, related media, links and alternate URLs. Equivalent to the `get_article` MCP tool. Credits: 1. Pass estimate=true to get the exact range without running the call, or max_credits to refuse anything costlier.

id — The article id, from any search result.

Search responses carry a bounded subset of an article's metadata, so a page of results stays a reasonable size even when one article contains fifty quotations. This endpoint returns everything.

The id is whatever a search result gave you. It begins with the article's collection timestamp (YYYYMMDDHHMMSS), which is how the sharded index routes a lookup to exactly one shard — so this is a cheap request regardless of how large the archive is.

Response

Everything an Article carries, plus:

Field	Type	Description
activity_density	number	Density of active language, as extracted.
authors	string[]	Bylines as extracted; pass one back as the `author` filter.
cities	string[]	Mentioned cities; pass one back as the `city` filter.
collected_at	string	When the index saw the article, 14-digit YYYYMMDDHHMMSS.
countries	string[]	Mentioned countries, canonical two-character identifiers; pass one back as the `country` filter.
event_types	string[]	Event classes (events-v1) implied by the article's GKG themes; pass one back as the `event_type` filter. Precise rather than complete: an article can describe an event without carrying its class.
id	string	Stable article id; pass to /doc/{id} and /similar/{id}.
image	string	Lead image URL, when the publisher declared one.
language	string	Source language, ISO-639-3.
locations	string[]	Mentioned place names; pass one back as the `location` filter. /doc/{id} has the structured form.
matched_locations	MatchedLocation[]	Present when near or bbox filtered the request: the mentioned cities that matched, nearest first.
mentioned_dates	string[]	Dates referred to in the text (YYYY, YYYY-MM, YYYY-MM-DD or --MM-DD); pass one back as the `mentioned_date` filter.
negative_score	number	Share of negative language, as extracted.
organizations	string[]	Organization entities as extracted; pass one back as the `organization` filter.
persons	string[]	Person entities as extracted; pass one back as the `person` filter.
polarity	number	Emotional charge regardless of direction, as extracted.
positive_score	number	Share of positive language, as extracted.
published_at	string	14-digit YYYYMMDDHHMMSS, UTC.
publisher_country	string	The publisher's canonical two-character country identifier.
regions	string[]	Mentioned regions (states, provinces); pass one back as the `region` filter.
score	number	Ranking score of this hit: BM25 for keyword, fused similarity for semantic, reciprocal-rank fusion for hybrid. Comparable within one response only; absent on a recency browse.
self_reference_density	number	Density of self-referential language, as extracted.
signals	Signals	Versioned signal intensities. Absent when the article predates the signal vocabulary.
source	string	Publisher domain, e.g. bbc.co.uk.
source_type	string	web \| citation \| academic_archive \| defense_archive \| journal_archive \| non_textual \| other.
story_id	string	The cluster this article belongs to. Pass to /stories/{id} to expand the story, or back as the `story_id` filter.
title	string	Headline as published.
tone	number	Overall tone, negative to positive, as extracted.
topics	string[]	Canonical topic identifiers; pass one back as the `topic` filter.
url	string	Canonical article URL. Link this; the publisher gets the visit.
word_count	integer	Words in the article body, when known.
amounts	Amount[]	Numeric amounts with the object they describe.
amp_url	string	The AMP version of the article, when declared.
date_mentions	DateMention[]	Dates referred to in the text, with precision. Distinct from `published_at`.
links	string[]	Outbound links from the article body.
location_details	ArticleLocation[]	Every mentioned place with its kind, country, identifiers and coordinates.
mobile_url	string	The mobile version of the article, when declared.
names	string[]	Proper names as extracted; pass one back as the `name` filter.
quotes	Quotation[]	Quotations as extracted, with the verb that introduced them.
related_images	string[]	Further image URLs from the article.
social_images	string[]	Images declared for social sharing.
social_videos	string[]	Videos declared for social sharing.

What the extra fields are for

quotes — extracted quotations with the verb that introduced them, each {text, verb}. Useful for "who said what" without fetching and parsing the article yourself, and filterable in search via quote_verb.
location_details — the same locations as the search response, but structured: each {country, kind, latitude, location_id, longitude, name, region_id, subregion_id}, with the kind (country, region, city), coordinates, and a canonical identifier you can filter on with location_id.
date_mentions — dates mentioned in the text, each {date, precision}, where precision says how much of the date the text gave. Distinct from published_at: an article published today can be about a date next year.
amounts — numeric amounts with the object they describe ("20000 combat soldiers"), each {number, object, value}. value is the number as written, a string that keeps the source's precision. number is the same amount parsed, or null when it does not parse. Filterable by what it counts with amount_object, and by number with amount_min and amount_max, where one amount must satisfy every bound.
amp_url, mobile_url, links — alternate and outbound URLs, for a fetcher that needs a lighter page or wants to follow the article's own citations.

Example

curl -s "https://api.unzoi.com/doc/20260826073000-a1b2c3" -H "x-api-key: $UNZOI_KEY"

An id that does not exist returns 404 with code: not_found — a statement of absence, safe to cache, unlike a 503. Article bodies are never returned by any endpoint — the index holds metadata and derived signals, and url is where the text lives.