Wikisignals.org: Helping Editors Evaluate Reference Reliability

Wikipedia’s credibility relies on its sources. To help editors reliably evaluate external references, a group of us at WikiCredCon 2025 built WikiSignals.org, a searchable database of website credibility indicators. Our goal is to include the trove of reference-reliability data in Wikimedia. But their lack of structured data and accessibility is an obstacle. This is our appeal to the community for help helping you. We need the community’s assistance with four proposals:

1. Wikimedia: Create a standardized, machine-readable template for Reliable Sources pages.
2. Wikipedia: Make external-link count by website domain easy to machine-harvest.
3. Wikidata: Populate missing news-source data in Wikidata, especially from Wikipedia article Infoboxes. 
4. Wikidata: Extend the news-source related properties in Wikidata.

Reliable/Perennial Sources

Wikipedians have spent 25 years defining reliable sources. Only 15 language projects (out of 347 active projects as of July 2026) have their own Reliable Sources Page (RSP). These lists have little coordination. Their formats vary wildly. At least six tools harvest the RSPs, including the popular Cite Unseen and Unreliable Source Detector. We each have our own parsing methods and frequency. This is inefficient.

A shared, standardized, consensus-based, and machine-readable template for RSP data would go a long way for projects in other languages to have their own RSPs. A Phabricator ticket related to this proposal is already underway.

Further, it would give language projects more visibility in how others classify domains. Currently, domains spam-blocklisted in one language are used in thousands of external links in another, a problem we suspect is partly driven by the lack of a shared resource.

External-link counts

We view a website’s external-link count — the total times its domain appears in the References, Citations, Notes, or External link sections — as a critical signal of credibility, proportional to Wikipedians’ trust in that site.

Currently, three services return a domain’s link count: a search form, the Quarry SQL tool, and an API. The language-agnostic reference risk model gives a related count of how many pages cite that domain. However, each has limitations: the search can’t limit results by namespace (e.g., articles only), Quarry lacks API access, the API’s pagination loops hit limits for high-count domains (complicated further by miser mode), and the reference risk model has not been updated since 2024.

At present, we rely on a kludged combination of API and Quarry results for an incomplete count. We propose creating an easily accessible way to get a domain’s external-link count totals, for a specific language and for all languages. Community wishes to add and update that data are included in the Community Wishlist.

Wikidata

Wikidata has more than 92,000 entries for news-related organizations. Currently, though, most of those entries lack critical credibility-related data.

Some of that missing Wikidata is in Wikipedia Infoboxes. The community needs an effort to populate this information in Wikidata, and existing information from Wikipedia article Infoboxes is one place where we can start.

We’ll be uploading WikiSignals’ wealth of data (on 70,000+ domains) into Wikidata in the coming months. However, there’s no place to store a key type of information. Adding this property would help Wikimedians evaluate reference reliability: 

• ‘indexed by’ — We need a property to indicate that a news source is included in a vetted list of legitimate media, such as Global Project Oasis. This is a valuable reliability signal, but we currently have no way to connect the index with the indexed item. Our request for comment is on Wikidata.

We’re not experienced Wikimedians and need assistance shepherding these new-property proposals. We can share our work harvesting RSP data and external-link counts. Please join us: Help us help you.