---
title: "GDELT DOC 2.0 API for Company News: Free, 15-Minute Updates, Limits, Independently Benchmarked (2026)"
description: "What the free GDELT DOC 2.0 API returns for a company query, its query syntax, update cadence and record limits, and where it falls short of a paid search API on company news."
url: https://openbenchmarks.com/company-news/gdelt-api
---

[](https://openbenchmarks.com/) 

[← company news benchmark](https://openbenchmarks.com/company-news) 

[benchmarks](https://openbenchmarks.com/) / [company news](https://openbenchmarks.com/company-news) / gdelt api 

not benchmarked · free and open · 15-minute updates

# GDELT DOC 2.0 API for company news

GDELT's DOC 2.0 API is a free, keyless search over the project's global news monitoring, updated every 15 minutes, returning up to 250 article records per query. It has not been run on the company-news benchmark. This page covers the query syntax, the limits, and why a free API that covers the world is still a hard fit for a question about one company.

The [benchmark](https://openbenchmarks.com/company-news)measures whether an API's top ten results contain the answer to a factual question about a named company. GDELT can be asked the question, but what it returns is a different shape from everything on the board.

Benchmark last measured 21 Oct 2026. 300 questions.

\[01\] the api

## Query syntax and what comes back

The endpoint is `https://api.gdeltproject.org/api/v2/doc/doc`. `query` takes keywords and quoted phrases, OR, and operators including `domain:` to restrict to a publisher. `mode=artlist` returns articles; the other modes return timelines, tone charts and image collages that are not useful here. `format=json`. `timespan` takes a count of minutes, hours, days, weeks or months, minimum 15 minutes; `startdatetime` and `enddatetime` take `YYYYMMDDHHMMSS`. `maxrecords` defaults to 75 and caps at 250\. `sort` is `HybridRel` by default, with date and tone orders available.

An article record is a URL, a title, the source domain, the language, a seen date and sometimes an image. There is no snippet and no body. That is the detail that matters for company news: if the fact is not in the headline, the API has not given it to you.

Coverage is the strength. GDELT monitors news in dozens of languages and the archive runs from 2017, so a company that is only covered in the regional press of one country is more likely to show up here than in an English-first index.

\[02\] cost and limits

## Free, with the limits that come with free

* **Price.** None. No API key, no plan.
* **Quota.** None published. The service is shared and throttles under load; there is no SLA and no support contract.
* **Records.** 250 per query, 75 by default.
* **Window.** Last three months by default; the archive reaches 1 January 2017.
* **Freshness.** Streams update every 15 minutes.
* **Company filter.** None. The name is matched as text.

Vendor prices, limits and parameter names were read from the linked pages on 18 November 2026. They are the vendor's list terms, not measurements.

\[03\] against the board

## What free costs on the lookup task

The benchmark has one free row: TinyFish, which answers 92.0% of the questions within its rate limit. The top paid row, Exa (type=fast), answers 99.3%. Both return snippets, which is how the extraction step finds the answer.

### Top five measured web search configurations on company-news questions

| Rank | Provider | Accuracy | Answer recall @5 | Mean search time | $ / 1k queries | $ / 1k correct |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | [Exa (type=fast)](https://docs.exa.ai/reference/search) POST /search · type=fast | **99.3%** | 99.3% | 652ms | $7.00 | $7.05 |
| 2 | [Context](https://docs.context.dev/search/overview) POST api.context.dev/v1/web/search · numResults=10 · highlights maxCharacters=1200 | **98.0%** | 98.3% | 15.93s | $4.40 | $4.49 |
| 3 | [Exa (type=instant)](https://docs.exa.ai/reference/search) POST /search · type=instant | **97.7%** | 97.3% | 398ms | $7.00 | $7.17 |
| 4 | [Perplexity](https://docs.perplexity.ai/api-reference/search-post) POST /search · search\_context\_size=low | **97.3%** | 98.3% | 1.38s | $5.00 | $5.14 |
| 5 | [Linkup (depth=fast)](https://docs.linkup.so/pages/documentation/endpoints/search/overview) POST /v1/search · depth=fast · outputType=searchResults | **96.7%** | 94.7% | 1.57s | $5.00 | $5.17 |

GDELT would need a fetch step to compete on this task, and a fetch step is exactly what the benchmark holds back from every vendor so that the comparison is search against search. Running GDELT with one is a separate experiment, and it would tell you about GDELT plus your scraper rather than about GDELT. For a researcher counting coverage of a company across countries and years, none of that matters and GDELT is the right tool. For an application that needs the fact now, the board is the answer.

[All news APIs and search APIs in one table →](https://openbenchmarks.com/company-news/best-company-news-api) 

\[04\] faq

## Questions answered on this page

### Is the GDELT API free?

Yes. The GDELT project describes its database as 100% free and open, and the DOC 2.0 API has no key, no plan and no published per-user quota. It is shared infrastructure run by a research project, so treat it as best effort rather than a contracted service.

### How fresh is GDELT?

The project says its streams update every 15 minutes. The DOC API's timespan parameter accepts a minimum of 15 minutes, which matches. By default a query covers the last three months; startdatetime and enddatetime narrow it, and the archive searchable through the API reaches back to 1 January 2017.

### How do I query GDELT for a company?

Call api.gdeltproject.org/api/v2/doc/doc with query set to the company name, mode=artlist for a list of articles, format=json, and a timespan such as 1week. The query supports phrases in quotes, OR, and operators such as domain:. There is no entity identifier; the name is matched as text, so a company with a common name returns noise and a company usually referred to another way is missed.

### What are the limits?

maxrecords returns 75 articles by default and at most 250\. A date range must fall within the archive window. The project states no request rate, and in practice heavy use gets throttled. Articles come back as URL, title, source, language and a few fields, with no body text, so the fact you are after has to be in the headline or fetched separately.

### Has GDELT been benchmarked?

No. Nothing prevents it: the API is free and the 300 questions fit inside a few hundred calls. The reason it has not been run first is the response shape. The benchmark scores whether the returned titles and snippets contain the answer, and GDELT returns titles without snippets, which puts it at a structural disadvantage the score would not distinguish from poor retrieval. A fair run needs a decision about fetching article text, which the benchmark does not do for anyone else.

\[05\] related pages

## Related pages

* [Company News Search Benchmark: full board and method →](https://openbenchmarks.com/company-news)
* [News API vs web search API →](https://openbenchmarks.com/company-news/news-api-vs-web-search-api)
* [Best company news API](https://openbenchmarks.com/company-news/best-company-news-api)
* [Event Registry API](https://openbenchmarks.com/company-news/event-registry-api)
* [NewsAPI.org pricing](https://openbenchmarks.com/company-news/newsapi-pricing)
* [Benchmark code on GitHub ↗](https://github.com/openbenchmarks-labs/company-news) · [Public dataset ↗](https://huggingface.co/datasets/openbenchmarks/OB-News-Websearch)

Openbenchmarks open benchmarks 

[company news benchmark](https://openbenchmarks.com/company-news) [news index APIs](https://openbenchmarks.com/company-news/news-index-apis) [web search benchmark](https://openbenchmarks.com/web-search) [source ↗](https://github.com/openbenchmarks-labs/company-news) © 2026 Openbenchmarks
