gdeltproject.org

What is The GDELT Project?

The GDELT Project is a free, open data platform that monitors global news media and turns it into a structured, queryable record of world events. According to its own description, it watches broadcast, print and web news from nearly every country, in over 100 languages, and identifies people, locations, organizations, themes, sources, emotions, counts, quotes, images and events. Its archives reach back to January 1, 1979 and update every 15 minutes.

What it actually is

GDELT is best understood as two things at once:

• A database: a machine-readable record of events, actors, themes and emotions extracted from news coverage.
• A graph: connections among those events, people, organizations and places over time.

The project frames its purpose as "computing on the entire world" — using large-scale computation to see global society through the lens of many media systems rather than a few Western outlets.

Who uses it, and for what

Practical trade-offs

The breadth is the point: coverage in over 100 languages and a decades-long archive let you study places and periods that narrower datasets miss. The trade-off is that this is derived data. Events, emotions and actors are extracted automatically from news text, so precision varies by language, source and story type, and the archive reflects what was published, not an independent record of what happened.

A useful next step

If you want to test whether GDELT fits your question, start with a narrow, checkable case — for example, one country and one month — and compare what the data reports against the original articles. That tells you quickly whether the extraction quality is good enough for your purpose before you commit to a larger analysis.

How can I access and download GDELT data for my own research?

The GDELT Project is built for exactly this purpose: it describes itself as a free, open platform for computing on global news media, and its data is meant to be queried, analyzed, and downloaded rather than viewed only through a dashboard. For most research uses, you start by choosing which layer of GDELT you need.

What you can download

• The event and network data. GDELT's core output is a machine-readable record of events, people, locations, organizations, themes, sources, emotions, counts, and quotes extracted from news coverage. The site frames this as "Computing on The Entire World: Events & Networks."
• The historical archive. The page states that its archives stretch back to January 1, 1979, which makes it viable for long-run longitudinal work, not just current-events monitoring.
• The realtime stream. Updates arrive every 15 minutes, so you can build a pipeline that refreshes on a schedule rather than downloading once.
• The blog. The GDELT Blog is the place to look for documentation-style posts and worked examples of what the data can do.

Practical routes in

1. Go to the "Querying, Analyzing and Downloading" section first. It is the page's own signpost for researchers, and it will route you to the access method that matches your skill level.
2. Use the "Get Started!" entry point if you want a quick look at current output before committing to a download workflow.
3. Decide between bulk files and queries. Bulk downloads suit reproducible research where you control the full dataset; query interfaces suit exploratory questions where you only need a slice. GDELT supports both, so pick based on whether you need the whole corpus or a filtered subset.
4. Plan for volume. A firehose updating every 15 minutes across more than 100 languages produces a lot of records. Store it in a database or columnar format rather than spreadsheets, and version your snapshots so results stay reproducible.

A concrete scenario

Suppose you are studying how a regional conflict is covered outside Western media. You could pull the event records for the relevant countries and date range, join them to the source and language fields, and compare tone and volume over time. GDELT explicitly positions itself as moving beyond a Western-media focus, which is what makes this kind of comparison possible from a single source.

Trade-offs to weigh

The breadth is the appeal and the limitation. Automated coding gives you scale and consistency across 100-plus languages, but it will not match the nuance of hand-coded content analysis, and you should expect to clean and validate before drawing conclusions. The archive depth back to 1979 is unusual, but older coverage reflects the media landscape of its time, so treat long time series with care.

Where to look next

For the underlying news sources and their own archives, the original publishers are often the better check on a specific article. Useful official domains include Reuters, Associated Press, and BBC. For the data itself, start at The GDELT Project's downloading section, pick one narrow question, and pull a small slice before scaling up.

What types of events and entities does GDELT extract from news articles?

GDELT extracts a structured layer of people, places, organizations, themes, emotions, counts, quotes, images and events from news coverage. It is designed to turn unstructured journalism into data that can be queried, analyzed and downloaded, with updates every 15 minutes and historical archives reaching back to January 1, 1979.

Core event and entity types

• Events — actions and interactions reported in the news, such as cooperation, conflict, diplomacy, protests, violence, agreements and other developments. These are the central unit for tracking what happened, where and when.
• People — individuals named or referenced in coverage, useful for tracking figures across countries and languages.
• Locations — geographic references, from countries and cities down to more local places, which allow event mapping and regional comparison.
• Organizations — companies, governments, parties, militaries, NGOs and other groups involved in or mentioned alongside events.
• Themes — recurring topics and issue categories, such as conflict, economy, health, environment, migration and human rights.
• Emotions — sentiment and emotional tone expressed in coverage, which helps distinguish how events are being framed rather than only what happened.
• Counts and quotes — numeric references and direct quotations, adding quantitative and rhetorical detail to the event record.
• Images — visual material associated with news items, extending the dataset beyond text.

What this means in practice

A researcher studying food security could combine locations, themes and counts to see where shortages are being reported and how coverage changes over time. A newsroom analyst could track organizations and people around a breaking story to identify who is being linked to an event. A conflict-monitoring team could use events and emotions together to spot rising tension in a region before it dominates international headlines.

The trade-off is that GDELT is broad and near-real-time rather than deeply curated. Its value comes from scale, language coverage and open access, not from perfect precision on every individual record. For a concrete next step, start with one country and one theme, then compare event counts and emotional tone across a short time window to see whether the signal is strong enough for your question. For related global news monitoring, see The GDELT Project.

How does GDELT differ from other global news databases or event datasets?

GDELT differs from most global news databases and event datasets in three practical ways: it is built as a continuously updating open data stream rather than a searchable archive, it codifies news coverage into structured events and networks rather than leaving it as text, and it aims for near-global language and geographic coverage instead of a Western-media focus.

What that means in practice

• A firehose, not a reading room. Many news databases are designed for finding and reading articles. GDELT is designed for computational analysis: it monitors print, broadcast and web news, extracts people, locations, organizations, themes, sources, emotions, counts, quotes, images and events, and exposes the result as open data. That makes it a better fit for quantitative research, monitoring dashboards and machine-learning pipelines than for simply browsing stories.
• Events and networks as first-class data. Datasets such as conflict-event catalogs often focus on a specific domain, such as political violence. GDELT's scope is broader: it attempts to represent global society through events, networks and emotional tone, which supports questions about discourse, reactions and relationships, not just incidents.
• Coverage beyond Western media. GDELT monitors news in over 100 languages from nearly every country, with archives going back to January 1, 1979, updating every 15 minutes. That matters if your question concerns local events or perspectives that rarely appear in English-language wire coverage.
• Open access. The platform is described as a free, open resource for research, which lowers the barrier for academic, civic and journalistic projects that cannot pay for commercial news APIs.

Trade-offs to weigh

The breadth that makes GDELT distinctive also creates limitations. Automated coding of multilingual news can introduce errors, and event counts reflect what media report, not an independent record of what happened. Coverage density varies by country and language, so comparisons across regions need care. For a narrow, carefully curated question, a domain-specific event dataset or a licensed news archive may be more precise. For large-scale, cross-language, continuously updating analysis, GDELT's design is the differentiator.

A useful next step

Start with a bounded question and a defined time window, then download the relevant GDELT data and compare a sample of coded events against the underlying articles. If the coding holds up for your use case, scale out; if not, consider a narrower dataset for the specific domain you need. The project's own getting-started material and blog are the natural entry points, and for complementary perspectives on news coverage you can compare against outlets such as Reuters or Associated Press.

What tools or APIs does GDELT offer for querying and analyzing its data in real time?

GDELT is built as an open data firehose rather than a single packaged analytics product. Its querying and analysis tools are mostly ways to pull from the same continuously updated stream of global news data, then process it yourself or in a cloud environment.

Core access points

• BigQuery: GDELT's datasets are exposed through Google BigQuery, which is the main route for large-scale, SQL-style querying over the full archive. This suits analysts who want to join GDELT events with their own data and run repeated queries without downloading everything.
• Raw file downloads: The project publishes downloadable data files that update frequently, so you can ingest them into your own database, pipeline or notebook. This is the most flexible option and the one most researchers use for reproducible work.
• The GDELT Blog and documentation: These provide the schema explanations, field definitions and worked examples needed to interpret the event, theme, emotion and network data correctly.
• Visualizations and "The World Right Now, This Moment": These are entry points for seeing what the data looks like before committing to a technical integration.

What "real time" means here

GDELT states that its archives update every 15 minutes and stretch back to January 1, 1979, monitoring print, broadcast and web news in over 100 languages. So "real time" is near-real-time in 15-minute increments, not a streaming API with sub-second latency. Plan your analysis around that cadence.

Choosing between them

A practical starting scenario

A researcher tracking protest activity across several countries would typically download the event files, filter by event type and geography, then cross-check the theme and emotion fields before drawing conclusions. The 15-minute refresh makes it viable for monitoring emerging situations, but the language coverage and coding process mean you should validate against local sources before acting on a single signal.

For related official sources, see The GDELT Blog and Google Cloud for BigQuery access details.

What are some practical applications or research use cases of GDELT data?

GDELT is most useful when you need a broad, continuously updated picture of what is being reported around the world, rather than a narrow or curated dataset. Its practical value comes from turning global news coverage into structured signals—events, actors, locations, themes, emotions and counts—that can be filtered, compared and analyzed over time. That makes it a strong starting point for monitoring, exploration and hypothesis generation, though not a substitute for careful source verification.

Common research and application areas

• Conflict and protest monitoring: Track where and when clashes, protests, repression or cooperation appear in news coverage, then compare trends across countries or regions. The historical depth back to 1979 is especially useful for long-run comparisons.
• Diplomacy and international relations: Map interactions between governments, leaders and organizations—who is meeting, praising, criticizing or sanctioning whom—and how those relationships shift.
• Media and framing studies: Compare how different countries or outlets cover the same event, including tone, emotional language and the themes they emphasize.
• Humanitarian and early-warning work: Use spikes in reports about displacement, disasters, violence or unrest as a prompt for deeper investigation by analysts or field teams.
• Economics and supply-chain awareness: Watch coverage of strikes, port disruptions, regulatory changes or commodity-related unrest that may affect markets or operations.
• Public health and social behavior: Track news volume and sentiment around outbreaks, vaccination campaigns or health policies across languages and regions.
• Computational social science: Build network graphs of people, organizations and locations; study emotion and discourse; or test theories about how narratives spread across borders.

A practical scenario

A small policy research team wants to know whether reported tensions between two countries are rising. Rather than reading a handful of English-language outlets, they query GDELT for events involving both countries over the past year, filter by event type and tone, and chart the weekly volume. A sustained increase in negative interactions becomes a signal to investigate specific incidents in the original reporting. The team treats GDELT as an index of coverage, not as ground truth: it tells them where to look, and they then verify with primary sources.

Trade-offs to keep in mind

GDELT's breadth is also its limitation. News coverage is uneven—some regions and languages are better represented than others—and the automated coding of events, actors and emotions can introduce errors or ambiguity. A single dramatic story can dominate counts, and the same event may be reported many times. For rigorous work, pair GDELT with a defined research question, transparent filtering rules and validation against a sample of source articles.

Getting started

Begin with a narrow question and a specific time window, then explore the querying, analysis and download options described on The GDELT Project. For related open-data and computational social science resources, you might also look at The GDELT Project if you need the same platform's project pages. A useful first step is to pick one event type and one country, chart it over several months, and manually check a dozen underlying articles to understand how well the coded data matches the reporting.