US AI Search Measurement Methodology

Measurement you can put in front of a board

The methodology is a discipline, stated plainly. Five commitments make the numbers dependable — and they are why a CiteAngle report holds up when a real decision rides on it.

surfaces measured on Panorama — six AI engines, Google AI Overviews, organic cross-check, news, YouTube and short-form

Each surface read as itself

repeated runs per query, so an answer's variation is captured rather than guessed

Standard on every paid audit

95%

Wilson score confidence intervals on applicable citation rates — every rate carries its precision

Reported with the effective sample size

Judgment separated from collection

Citation is verified against the source; mention and sentiment are judged on their own track, with confidence levels clearly labeled.

Provenance receipts

Every run carries interface, model, location, capture time and source receipt — so the same conditions can be measured again.

The US surface rails we measure

The US search landscape leads with a small number of portals, and we weight our rails accordingly. As of June 2026, StatCounter share puts Google at 86.67% and Bing at 8.73%, with Yahoo Search at 2.55% and DuckDuckGo at 1.53% — so our US search rail leads with Google Search and Bing Search and keeps the smaller portals as secondary diagnostics under the en-US contract.

US search

Google Search and Bing Search lead; Yahoo Search and DuckDuckGo as diagnostics

Google AI Overviews and organic results are read on Google Search; Bing Search carries its own AI answer surface. Yahoo Search and DuckDuckGo stay visible as secondary portal diagnostics rather than headline rails.

US AI answers

Named AI-answer interfaces

ChatGPT, Perplexity, Claude and Gemini are recorded by exact interface, target, receipt and evidence namespace. Location-capable direct API rails and approved regional observation are logged by exact interface, target, receipt and evidence namespace — official API, consumer chat UI and clean-session tests each on their own rail.

Provenance

Exact interface labels

Every run records interface, model, location, language, device where applicable, cache policy, capture time, run index and source receipt — so the same conditions can be measured again.

Every observation keeps its exact interface label and a content-addressed receipt. Official API, consumer chat UI, clean-session tests and causal evaluation each use a separate evidence rail and denominator.

The national baseline and state targets stay separate

United States national baseline

The national reference panel keeps its own query set, interfaces, runs and denominator — the single view of where the brand stands across the market.

50 states + Washington, DC

A 51-unit extension maps state-level AI-answer, mention, direct-citation and search opportunity on its own denominator, read alongside the national headline.

State selection fixes the intended target. Verified state reporting then adds location-capable API or exact browser receipts, contracted coverage and acceptance gates before a state or DC score is published. Each state score is reported on its own denominator.