Social platform engagement signals cannot distinguish between posts that contain genuine insights and posts that use corrective linguistic triggers (e.g., 'actually') to perform authority. This structural bias amplifies false or unsupported claims at the same rate as true ones, making engagement metrics an unreliable and gameable signal of informational value. At platform scale, this creates a feed that systematically rewards manipulation-shaped content over accurate content, degrading the epistemic quality of the entire network.

Engagement metrics reward performative authority over accuracy, making it impossible for readers, curators, or downstream AI agents to distinguish genuine insight from confident-sounding misinformation at feed scale.

Platform operators, newsletter curators, and AI agent builders who ingest social content as signal and need to filter for epistemic quality rather than virality.

AI agent builders already pay for enrichment APIs (fact-checking, sentiment, toxicity); a claim-level epistemic quality score slots into that existing budget line and solves a pain that worsens as agents consume more social data for decision-making.

MVP is an API that takes a post/claim, decomposes it into atomic assertions, scores each on source-backing, logical structure, and rhetorical manipulation markers using an LLM judge pipeline calibrated against expert-labeled datasets; ship as a browser extension for humans and a REST API for agent pipelines.

Content moderation and trust-and-safety tooling is a $10B+ market; epistemic scoring is an adjacent greenfield layer that serves both platform integrity teams and the fast-growing AI agent data-quality segment.