MWITA-ASDO-2026-005 · Evidence B · P1
A defined AEO intervention on one large site's treated page subset produced a 1.82-fold intervention-aligned increase in the treated/control ChatGPT-referral ratio, far below the site's raw 5.7-fold referral growth.
What this does not establish
Single-domain bundled intervention cannot isolate which tactic worked or establish a portable revenue effect; ChatGPT platform growth and a short noisy pre-period limit attribution.
Counterevidence & uncertainty
Total referrals grew 5.7x but untreated pages grew 3.5x, showing that raw before/after multiples substantially overstate intervention incrementality.
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.