Kevin Indig Finds 91% of AI Citations Appear on One Platform

According to Indig, tracking a brand's presence in AI search requires accounting for the engine, personalization, reasoning levels, model updates, and stochastic variability. His report says the citation and mention overlap across engines is small; Search Engine Journal's August 8 analysis highlights the same 91% figure while treating the report as the source of the finding.

A conventional search workflow often treats rankings as comparable observations across a known results page. Generative answer systems can instead select different sources and produce different answers for a similar query depending on the platform and execution context. The reported finding does not establish which engine is most accurate or valuable for a particular brand, but it does show that a single-platform observation is incomplete evidence of cross-platform visibility.

Indig also argues that attribution is the common issue across several H1 2026 AI developments. The report cites uncertainty around measuring AI visibility, connecting inference spending to return on investment, distinguishing perceived from actual software disruption, and identifying the causes behind layoffs attributed to AI.

For data and marketing analytics teams, a defensible measurement design would record more than a rank or binary citation outcome. Useful fields include the engine, prompt formulation, collection time, observable model or product version, geographic and personalization conditions, cited domains, and downstream conversion events. Repeated sampling cannot remove stochasticity, but it can expose variation that a one-time answer capture hides.

Teams using the 91% figure as a planning input would still need their own prompt set, platform coverage, sampling cadence, and outcome definitions before translating the finding into channel or content decisions.