From language to coordinates An embedding represents contextual meaning as a coordinate. Related ideas cluster together, while unrelated passages land farther apart. Cosine similarity measures the direction between those coordinates so semantic relevance can be compared even when two passages use different words.
See where content drifts The tool compares each target-content embedding with an anchor coordinate, such as your intended topic or the page's primary cluster centroid. Greater distance means greater orbital drift, revealing peripheral chunks that can dilute topical focus or send mixed relevance signals.
Realign without flattening Inspect drift chunk by chunk, edit the target or anchor content, and measure the geometry again. You can strengthen weak connections, remove genuine tangents, and preserve useful supporting ideas while moving the page closer to the semantic space you want it to occupy.
