Treewalker systematically analyzes a brand's presence by running paired ungrounded and search-grounded queries across multiple runs in Vertex AI. It inspects token log-probabilities to quantify the model's confidence for each identified entity, following alternative branching paths when alternate token probabilities clear a defined threshold. These ungrounded association trees are then cross-referenced against live grounded search outcomes to produce a holistic comparative gap analysis.