Data Collection Methodology

An overview of how fact-check records are sourced, validated, structured, and archived — from RSS feed to queryable dataset.


Monitored Feeds

The pipeline monitors internationally recognized fact-checking organizations. Content is fetched on scheduled intervals. To view a complete list of sources and their collection types, see the table below.

PolitiFact

RSS feed

U.S. political fact-checking from the Poynter Institute. Covers claims by politicians and public figures on a six-point rating scale.

Snopes

RSS feed

One of the oldest fact-checking publications. Covers viral claims, urban legends, and political misinformation across multiple countries.

FactCheck.org

RSS feed

Nonpartisan political fact-checking from the Annenberg Public Policy Center. Focuses on U.S. federal and state politics.

Lead Stories

RSS feed

Real-time fact-checking of viral misinformation using the Trendolizer platform. Covers health, politics, and social media hoaxes.

Full Fact

RSS feed

Independent fact-checking charity based in the UK. Covers claims in British politics, health, and public discourse.

Africa Check

Atom feed

Africa-focused fact-checking organization operating across multiple countries. Covers political claims, health misinformation, and social media content.

AFP Fact Check

RSS feed

Global fact-checking desk operated by Agence France-Presse. Covers viral and misleading content in multiple languages worldwide.


Three-Stage Ingestion Pipeline

The pipeline runs as a sequence of three independent stages with each stage producing a structured audit record regardless of outcome. Stages are designed to be deterministic and independently replayable so that archived content can be reliably reprocessed to extract new or updated information.

  1. 1

    Collection

    Fetch raw feed content and archive to durable storage. SHA-256 content hash serves as a deterministic lineage identifier.

  2. 2

    Sanitization

    Apply an access-control policy to classify the safety of each content item. Policy decisions are recorded in an audit action list.

  3. 3

    Extraction

    Parse feed entries from safe content, extract and normalize select fields and write structured data to temporary loading storage.


Data Integrity & Auditability

All requests for content — successful or failed, passing or quarantined — are archived for auditing purposes. No original content is overwritten.

We derive a deterministic content_lineage_id from collected content. Re-ingesting duplicate content at a later time produces the same identifier, enabling idempotent re-runs and exact-match deduplication. Because all policy decisions are recorded as ordered action lists in archived records, the full history of how any observation was classified is queryable and reproducible. Quarantined content is retained for manual review rather than discarded.

To protect the integrity of the data we collect, the fact check database encrypts archive data at rest and applies least-privilege access patterns throughout our technology.