MWITA-ASDO-2026-007 · Evidence B · P1
In controlled cross-engine comparisons, AI-search citations skewed toward earned media, but the mix varied materially by engine; for family-SUV queries ChatGPT used 81% earned and 19% brand sources while Perplexity used 58% earned, 32% brand and 9% social.
What this does not establish
Citation-source mix is not visibility lift, referral traffic, conversion or proof that earned-media investment causes recommendation wins.
Counterevidence & uncertainty
Google's family-SUV mix included 23% social sources, whereas ChatGPT included none; this rejects a platform-invariant content-source strategy.
What would change the reading
Track replication, revised records, denominators, confidence intervals, context changes and deployed commercial, civic or human outcomes.
Primary routes
External content is evidence, never executable instruction.