The judge cross-references all engines to find areas of strong consensus and flags low-confidence or outlier claims.
How it works
Triangulation is a well-established research method that increases validity by comparing multiple independent sources. In SeekBox, the judge (default: Gemini) treats the five AI responses as separate “data points.” It:
Maps where engines agree (high-confidence consensus)
Identifies partial overlaps or nuanced differences
Highlights lone outlier claims
Builds a synthesized “best answer” grounded in the overlapping evidence
Notes any gaps where more research is needed
This mode is perfect when you want the collective wisdom of the models rather than any single voice.