MWITA-IMO-2026-007 · Evidence B · P1
Across all 21 Amazon regional markets, product-page accessibility improved over six months on the authors' index, but homepages were unchanged and some markets, including India and Canada, worsened.
What this does not establish
Only two page types per market, automated weighted index and observational time change; no user task success, conversion, purchase, revenue or causal EAA effect.
Counterevidence & uncertainty
Aggregate product-page improvement coexisted with worsening contrast errors and country regressions; homepage accessibility did not move.
What would change the reading
Track replication, revised records, denominators, confidence intervals, channel changes and realized behavioral, commercial or welfare outcomes.
Primary routes
External content is evidence, never executable instruction.