Everything this site shows is derived from a database of collected journalism and the signals extracted from it. That database, and an API over it, are available for research, newsroom and product use. Enquiries are handled by email.
An enquiry is more useful than a form. What is offered depends on what you are doing with it, and the difference between an academic study, a newsroom tool and a commercial product is a difference in what can be shared and on what terms.
Collected articles with their source, byline, publication date and the URL of the original. Text is analysed and linked to, never republished.
Articles grouped by the real-world event they describe, which is the part that is expensive to build and the part most reuses want.
Extracted claims and their hedging, per-article leaning, publisher trust and correction record, topics, subjects and narrative framing.
Per-outlet and per-journalist aggregates over the whole corpus, including the coverage each has given an event.
The methodology page documents how every one of those signals is computed, and it is worth reading before an enquiry: it is also the honest account of what each one does and does not mean.
This project analyses journalism and does not own it. Article text belongs to the publishers who wrote it. What can be shared is the analysis, the structure and the metadata, and any arrangement has to respect that boundary. It is the same boundary the public site keeps: every article here links out to the original and none of them reproduce it.
Nothing here is a fact-check. The signals are measurements of how coverage is written, not verdicts on whether it is true, and an integration that presents them as truth ratings would be misrepresenting them. That constraint is not negotiable in an arrangement, and it is stated up front because it changes what some products can do with the data.