Impact
What Veridi enables
For the public
A structured, transparent way to evaluate claims that shows its reasoning. Rather than asking people to trust a verdict, Veridi shows the evidence, the source quality, the confidence level, and any manipulation techniques detected. Readers can evaluate the assessment’s quality for themselves.
For journalists and fact-checkers
A systematic second-opinion tool and gaming countermeasure scanner. Professional fact-checkers can use the methodology to check for disinformation techniques that are designed to evade detection; particularly framing manipulation, where every individual fact is true but the composite is false, and institutional capture, where formerly authoritative sources have been compromised.
The source hierarchy and decision trees also serve as training material, making implicit editorial judgment explicit and teachable for new fact-checkers.
For researchers
An open, documented methodology that can be studied, evaluated, and improved. The complete methodology files, validation data, and test suites are available for inspection. The adversarial testing framework - particularly the taxonomy of eleven disinformation attack vectors - may be independently useful for researchers studying information manipulation.
For institutions
A fact-checking process that can be audited. When an institution relies on Veridi for information verification, the assessment trail provides accountability: not “we checked and it’s true,” but “here are the sources (classified by quality), here is the reasoning (following this decision tree), here are the manipulation checks (these were applied), and here is the confidence level (bounded by these structural caps).”
Scale characteristics
The methodology is implemented as an AI prompt system, which means:
- Speed: A Standard-tier assessment takes minutes. A Full-tier assessment with specialist analysis takes longer but is still substantially faster than manual fact-checking.
- Consistency: The same claim submitted twice follows the same procedures. This doesn’t guarantee identical outputs (evidence availability changes and AI reasoning varies), but it ensures the same structural checks are applied.
- Cost: Operating costs are primarily AI inference costs, which are substantially lower than employing full-time fact-checkers per assessment. The lean nonprofit model keeps overhead minimal.
- Language: The methodology has been tested with non-English sources in Japanese, Turkish, Chinese, and Hindi. The AI implementation can evaluate sources in most written languages.
What Veridi does not claim
- It does not claim to be infallible. 96/97 is strong but not perfect, and self-tests have limits.
- It does not claim to replace human judgment. Editorial decisions about which claims to investigate, how to present findings, and how to engage audiences remain human responsibilities.
- It does not claim to solve disinformation. No single tool can. Veridi provides one layer of defense - a structured, auditable, adversarial-aware methodology - in a problem that requires many.
- It does not claim the AI is uniquely capable. The methodology is designed to be separable from any particular AI system and could in principle be followed by human fact-checkers or implemented on different platforms.