Agent traffic passes half of requests

For a typical newsroom, HTML requests from automated agents now rival human page views. The interesting part is not the volume; it is the mismatch between what agents fetch and what they need. Most pulls are re-reads of pages that did not change.

Signed indexes change that arithmetic. When digests are part of the contract, an agent can skip the page entirely and spend its budget where the reporting actually moved. Publishers keep a record of what was read, under which declared permission.