An open, reproducible, multi-signal domain credibility dataset for automated pre-bunking of online misinformation. Designed for on-device, client-side deployment.
CRED-1 is a domain-level credibility dataset combining multiple openly-licensed source lists with computed enrichment signals: source labels, domain age, web popularity, fact-check frequency, and threat intelligence. It provides credibility scores for 2,672 domains known to publish mis- or disinformation, conspiracy theories, or other unreliable content.
Unlike server-side credibility services, CRED-1 is small enough to ship inside an application as a static JSON snapshot, which keeps lookups fully client-side and preserves privacy. The dataset, scoring pipeline, and source attribution are open and reproducible.
A. Loth, M. Kappes, and MO. Pahl. CRED-1: An Open Multi-Signal Domain Credibility Dataset for Automated Pre-Bunking of Online Misinformation. Preprint, 2026.
The work is being presented at the 18th ACM Web Science Conference (ACM WebSci 2026) in Braunschweig, Germany.
Trackless Links is the first production integration of CRED-1. The iOS and macOS Safari extension uses the bundled CRED-1 snapshot to warn users about low-credibility destinations on-device, with no server calls and no telemetry. Tracker stripping and credibility lookup happen in the same pass before the destination site is contacted.
To mark the WebSci 2026 presentation, free promo codes for Trackless Links are available for readers and conference attendees while supplies last.