FactHarbor for fact-checkers and newsrooms
FactHarbor is a non-profit, open-source project for evidence-based claim analysis. It researches text and articles, separates verifiable assertions, and produces reports with supporting and opposing evidence, reasoning and uncertainty.
What a newsroom can inspect
Full-size diagram · Mermaid source
Evidence can support different conclusions under different conditions. FactHarbor's analytical boundaries make those conditions visible; its debate method challenges an assessment using documented evidence. The report should let a reader examine the sources and understand why a conclusion was reached.
This can support pre-publication research, identification of evidence gaps and teaching how an assessment is constructed. Editorial judgment, original reporting and specialist image/video verification remain separate responsibilities. An automated report is not a substitute for checking the cited material.
Alpha interface examples
These historical screenshots illustrate the report format, not a current performance benchmark.
Status and limitations
The invite-gated Alpha remains available at app.FactHarbor.ch. Broader development is paused pending funding. Reliability, repeatability, multilingual consistency and resource efficiency need further evaluation. No fixed turnaround time or uniform accuracy is promised.
A report may miss relevant sources, misinterpret evidence or express too much confidence. Truth assessment and confidence are distinct; weak or conflicting evidence needs visible qualification. Citations help scrutiny but do not guarantee correctness. Model allocation depends on the active configuration, so a static presentation does not establish which providers ran a particular analysis.
Possible cooperation
A useful first discussion would identify one workflow problem, what evidence would demonstrate improvement, and a manageable evaluation scope. Newsrooms could contribute workflow knowledge and assessment of report usefulness; FactHarbor can offer an Alpha application and inspectable outputs. Access, data handling, resources and publication arrangements must be agreed before a study or pilot starts.
Contact Robert Schaub at info@factharbor.ch. See academic cooperation, support and quality and trust.