Replace, Don't Erase: Corrective Substitution and Reach Inheritance as Structural Remedies for AI-Driven Health Misinformation

Background: Existing interventions against AI-driven health misinformation (removal, algorithmic deamplification, labeling, community annotation, and retrospective notification) differ in mechanism but share an unexamined premise: that the distribution infrastructure accumulated by a misinformation artifact, its URL, its position in reshare and reply graphs, its embedded instances, and its engagement history, is either damage to be destroyed, a hazard to be attenuated, or a fixture to be annotated. The correction literature has meanwhile documented a persistent failure, termed here the audience mismatch problem: corrective content rarely reaches the audience the falsehood reached. Objectives: This paper introduces a fourth relation between correction and accumulated reach: inheritance. It defines reach inheritance as a design principle and corrective substitution as its operationalizing mechanism, in which an identified misinformation payload is replaced, at its original locus, by an authorized corrective payload while every inbound pathway is deliberately preserved, so that the infrastructure built by the falsehood delivers its correction. Methods: Presented as a design-framework Viewpoint in continuity with the Multi-layered Epistemic Disruption Framework (MEDF), which identified content interception as an untested direction for design research, the paper synthesizes platform intervention research, the comparative media-law doctrine of the right of reply (droit de réponse, Gegendarstellung, and their American constitutional counterpoint), and the procedural precedent of notice-and-takedown regimes. Results: Corrective substitution is differentiated from seven adjacent intervention mechanisms, none of which both retires the false payload and transfers its distribution assets to the correction. A protocol specification is developed comprising a two-tier trigger architecture that concentrates the strongest modes on verifiable identity facts (clinician impersonation) rather than adjudications of truth; four substitution modes ordered by payload persistence; a five-stage authorization chain whose load-bearing act is the impersonated party's formal attestation of non-authorship; and a temporal decoupling of machine-speed provisional payloads from human-authored corrections. An adversarial analysis examines false-positive substitution, attestation fraud as an instrumentalized liar's dividend, governance capture, and the substitution pipeline itself as an attack surface, yielding explicit design constraints, scope conditions, and a minimal viable deployment confined to impersonation-triggered, tombstone-transparent substitution with pre-enrollment registries. Four testable hypotheses, doubling as deployment gates, are advanced. Conclusions: The intervention repertoire against AI-driven health misinformation is missing its fourth quadrant. Corrective substitution fills it as the platform-era completion of the right-of-reply tradition: the correction reaching, by construction, exactly the audience the injury reached. The mechanism is domain-specific, jurisdiction-dependent, and offered as one layer-indexed component of the MEDF's intervention mapping rather than a standalone remedy.