What is GDELT?

Analysis tools, raw downloads and documented BigQuery datasets offer ways to investigate reporting patterns around a geopolitical question. Event records describe extracted actors, actions and geography; the knowledge graph adds entities, themes and relationships in coverage.

These outputs reflect both what media report and how automated systems classify it. Duplicate articles, uneven attention and changing source availability can distort counts. A coverage spike is not a verified number of events or a direct probability signal, and a coded event alone does not establish a contract’s settlement source.

Who it is for

OSINT researchers investigating geopolitical narratives and reporting patterns.
Developers building news-monitoring and event-analysis workflows.

Features and coverage

Structured event records
Explore actors, actions and geographic context extracted from news reports.
Knowledge-graph research
Investigate entities, themes and relationships represented in media coverage.
Multiple data-access paths
Use public analysis tools, raw downloads or documented BigQuery datasets, checking each external platform’s requirements.

Workflows with GDELT

Investigate a reporting pattern

Define topic, location and time window, filter the records and inspect the original articles behind relevant results.

Establish a monitoring baseline

Keep queries and source selection consistent when comparing coverage, before interpreting an apparent increase.

Pricing and total cost

GDELT describes its database and public analysis services as free and open. External computing and BigQuery usage may incur separate charges.

Access: Public documentation and downloads are available; external analysis platforms have their own access requirements.

Getting started

Before you start: Automated extraction can misclassify material.

Select the event dataset or knowledge-graph dataset for the question.
Define a consistent topic, geography and date range.
Inspect original reporting before treating a record as evidence.

Public docs and downloads are accessible; any external analysis environment has separate access conditions.

Useful links

What to check

Automated extraction can misclassify material.
Duplicate coverage and uneven media attention can distort counts.
Verify material claims against original reporting and authoritative sources.
A research record is not automatically the settlement source named by a prediction contract.
External compute and BigQuery can add costs despite free public data.

A closer look

Ratings & reviews

Editorial rating

4.0/ 5

0 user reviews

PredictionToolkit editorial

Our take on GDELT

A broad open-data resource for event research and reproducible news analysis. Its automated classifications require checking the underlying sources before using them as evidence.

Based on documented features, setup and access terms. How we rate