The Comprehension Gap: Why AI Communication May Be Evaluated at the Wrong Point

This working paper proposes that AI-mediated communication systems operating in human-impact domains may currently evaluate success at the point of output while the conditions required for human comprehension are established upstream of generation. The paper introduces the concepts of the Pre-Deployment Failure Layer and the Comprehension Integrity Stack as exploratory frameworks for understanding comprehension-centered infrastructure in multilingual civic communication systems. Rather than positioning comprehension as a downstream quality concern, the paper argues that communication legitimacy, accessibility conditions, linguistic authority, and trust formation pathways may require structural evaluation before deployment. Awareness365 is referenced as an exploratory implementation environment for testing governance-centered communication architecture under multilingual real-world conditions.