JMIR Publications

ABSTRACT

Scientific misinformation is usually treated as a content problem: identify what is false, fabricated, or manipulated, then flag or remove it. This Viewpoint argues that it is also an information-architecture problem. The scientific record carries claims far more effectively than it carries evidence of how those claims were verified, by whom, and with what subsequent history. As a result, downstream readers, institutions, and increasingly machines cannot discriminate among records on verification grounds. The problem is sharpening as technology improves. Detection and attribution are different problems: as generative artificial intelligence (AI) collapses the cost of producing plausible scientific text and paper mills industrialize fabrication, detecting problematic content no longer establishes who is responsible for it, yet classical deterrence economics assumes that the second follows from the first. We propose Proof of Verification (PoV), an architectural proposal in which four properties of the scientific record (verification events, attribution, provenance, and publication state) become persistent, inspectable, and portable elements. PoV composes existing infrastructure (more than 1.1 million peer-review records already registered with Crossref; ORCID identity; C2PA and W3C PROV provenance standards) rather than creating a new platform. Plural validation is a safeguard against capture; institutional staking is strictly optional. PoV does not establish scientific truth; it makes the basis of trust claims inspectable. Its central theoretical proposition is a downstream legibility channel added to classical deterrence: verification state that is observable, understood, and used can reduce the value of unverifiable records to readers, institutions, search systems, and AI-mediated retrieval; the channel is empty wherever that state is ignored. We state the architecture’s failure modes (privacy, governance, gaming, capture, and Goodhart effects) and derive falsifiable predictions, two testable with existing data. Scientific misinformation controls should move beyond detecting problematic content toward making verification continuously inspectable.