Amazon's Rufus and Cosmo AI are upending keyword-based listing strategy

A 90-minute training session making the rounds among Amazon sellers lays out how the marketplace's AI systems, the shopper-facing assistant Rufus and the underlying Cosmo algorithm, are dismantling traditional keyword strategy.

The core claims: Rufus is now used by half of younger shoppers and growing, while Cosmo ranks products on inferred buyer intent rather than keyword matches. Amazon is also analyzing product images for ranking signals, effectively scanning visuals the way it once scanned text. One much-cited example: pregnant women searching for shoes surfaced as a distinct intent cluster, illustrating how deeply the system models life context.

The warnings are stark. Keyword-stuffed titles that once won placement are now said to hurt rankings, and sellers earning millions annually could see revenue collapse if they ignore the shift. A prediction attributed to Walmart's leadership, that search bars will disappear within two years, gets quoted as evidence of where all retail discovery is heading.

Practitioners describe 15 relationship types the Cosmo system uses to understand products, most of them unknown to sellers, and frameworks borrowed from Silicon Valley product management for prioritizing optimization work.

Tooling is following the knowledge: custom AI bots promising ten-minute listing optimization, image generation systems built to stop the scroll, and audit checklists for Cosmo readiness are becoming standard seller kit.

Whether or not the most dramatic revenue-loss predictions hold, the direction is consistent across sources: listings written for algorithms that read keywords are being rebuilt for algorithms that model shoppers.