Retrieval Basin Topology is a theoretical and measurement framework by Lee Sharks for studying how AI retrieval systems settle on competing framings of a term.
The paper describes a retrieval basin as a region of the public semantic field that retains and attracts a particular framing. Deposits, metadata, citations, user engagement, and generated summaries are treated as different strengths of writing into that field. A term with no serious competitor may stabilize quickly; a generic term may be pulled toward older and more authoritative scientific or commercial meanings.
RBT proposes several instruments. Retrieval Capital aggregates source mass, institutional authority, citations, age, and engagement. Basin Depth Ratio compares a target framing with its strongest competitor. Framing Persistence Index measures how often the framing appears across repeated tests. Drift Vector measures directional change, and Basin Capture Velocity estimates movement following deposits.
The work includes a state machine and operational query guidance. Its calibration is explicitly preliminary: the principal thresholds derive from two cases, hidden platform mechanisms are not observed directly, and platform-specific geometries are hypotheses based on public outputs and cited secondary literature. RBT is therefore a falsifiable research program and archive governance tool, not a settled science of search engines.