Skip to content

FactHarbor

Open-source applications and web services for AI-powered fact-checking — every verdict backed by evidence you can inspect.

Our Vision and Mission

Our Vision A world where decisions and public debate are grounded in evidence so people can move forward with clarity and confidence.

Our Mission The FactHarbor application brings clarity and transparency to a world full of unclear, contested, and misleading information by shedding light on the context, assumptions, and evidence behind claims.

Non-profit. Open source.

  • FactHarbor association (Verein): The Swiss non-profit organization behind the work. It serves the public interest with transparency and no profit motive.
  • FactHarbor application: The open-source fact-checking software developed within the association. Its sources, evidence and reasoning can be inspected.

FactHarbor application’s next chapter

FactHarborDSS is the planned successor application, extending fact-checking to decision support.

Read a FactHarbor report

Open FactHarbor application to explore reports. FactHarbor is an Alpha; access to new analyses is by invitation.

Learn how to interpret a FactHarbor report.

In the presentation example below, the first entry in the reports list is labelled Mixed, with Medium confidence. It concerns generative AI and the preparation needed for small-business automation. These are the Alpha report's assessments, not independently validated conclusions or DSS output.

FactHarbor Alpha reports list, captured on 24 October 2026

Reports list captured 24 October 2026. Open the image for a larger view.

Inspect the evidence

The evidence matrix separates parts of the broader claim and groups evidence with compatible conditions—such as dates, locations and methods—into assessment boundaries. Follow an item to its source and ask: Does it support the point? Does it apply here? What contradicts the conclusion, and what is missing?

Evidence matrix from the selected Alpha report, with separate subclaims and evidence counts

From the same example. Colours summarize the report's assessments; counts are evidence items, not votes or measures of evidence quality. Both screenshots are reproduced unchanged from the project presentation of 26 October 2026. Attribution: FactHarbor and its contributors.

Open the selected Alpha report. Availability and access depend on the application; these dated images illustrate the interface at that time.

How it works

The application identifies factual claims, researches supporting and opposing evidence, and assesses the findings. A structured agent debate challenges the assessment before reconciliation and validation; model roles depend on the active configuration. Evidence-backed challenges can change a conclusion; agreement alone does not establish correctness.

FactHarbor application analysis: understand the claim, research evidence, assess and challenge findings, then produce a report

Full-size diagram · Analysis pipeline

Explore further