This paper introduces AIVO Orbit, a continuous evidence intelligence system designed to address the remediation bottleneck identified in AI decision-stage measurement programmes. Building on the CODA methodology (WP-2026-01) and the upstream measurement gap analysis (WP-2026-05), AIVO Orbit automates the discovery, classification, and pipeline population of evidence sources required to correct brand inference position at the AI purchase recommendation stage. The central problem Orbit addresses is structural: AIVO Meridian's decision-stage diagnostic framework identifies filter gaps with precision - classifying the specific displacement mechanism, the verbatim criteria the model applied, and the competitor that captured the T4 recommendation. However, the remediation pipeline that follows requires evidence file construction: identifying qualifying source URLs, assessing authority weight, classifying evidence type, and articulating what each source proves against the displacement criteria. For a single brand with four filter gaps, this is manageable. For an agency managing thirty brands across fourteen filter types and four AI platforms, manual evidence construction becomes the rate-limiting step that prevents the remediation loop from closing. Orbit resolves this by maintaining a live evidence watch for each active filter gap, monitoring a stratified source universe weighted by platform-specific authority signals, and surfacing candidate sources with pre-populated claim statements and evidence classifications. The human approval role shifts from discovery and assessment to review and authorisation. We describe the Orbit architecture, the four-tier source authority model, platform-specific evidence profiles for ChatGPT, Gemini, Perplexity, and Grok, the integration map within the AIVO product stack, and the commercial implications of continuous versus manual evidence construction. We propose Orbit as a necessary complement to decision-stage measurement - the instrument that ensures diagnostic findings are actionable on a continuous basis rather than requiring periodic manual intervention.
Orbit et al. (Mon,) studied this question.