The Third Shelf: How Amazon's AI Recommends Products | Autopilot

Amazon's AI ignores your rank nearly two-thirds of the time.

We ran 1,963 non-branded queries through Amazon's AI shopping assistant and mapped all 13,177 of its recommendations against the search page. ~64% of picks sit outside the organic top-10 — and ads alone can't buy your way in.

Download the study — free, no email Get the 903-brand Library ↓

1,963 queries · 70,176 products · 13,177 AI picks · 903 brands indexed

~64% decoupled

of AI picks fall outside the organic top-10. Your rank dashboard can't see this shelf.

2.4% ad-only

Ads appear on 14.3% of picks, but 83% of those already rank organically. Ads ride winners.

903 brands indexed

The Deep-Dive Library maps 683 queries, page by page. Are you on the list?

The Third Shelf — Deep-Dive Library

903 brands indexed. Are you on the list?

Want your own term mapped — free?

We're mapping one search term per brand on LinkedIn. Comment your term on the launch post, connect with Christian, and get your map in 48 hours.

This is the market view. Want yours?

A custom deep dive maps your brand's position across the roughly 100 queries you actually compete on — AI picks, organic ranks, and ad share, side by side.

Your 100 core queries

The non-branded terms and AI prompts your category runs on — built from search volume, your catalog, and competitor coverage.

Three-shelf share

Where you stand on each: organic rank, ad presence, and whether the AI recommends you — versus the brands it picks instead.

The gaps

Queries where you rank but the AI passes you over — and what its picks have in common that your listings lack.

Tracked over time

The third shelf moves. Quarterly re-runs show whether optimization is winning you picks — market-share based, revenue-linked.

About the study.

The Third Shelf is, to our knowledge, the first large-scale public mapping of Amazon's AI shopping recommendations against the search results page: 1,963 non-branded queries, 70,176 products, 13,177 AI picks, captured in 2026. Full methodology in the white paper.

Published by Autopilot × The Order.

This study is independent research. It is not affiliated with, endorsed by, or sponsored by Amazon.com, Inc. Amazon, Rufus, and Alexa are trademarks of Amazon.com, Inc. or its affiliates. All product and brand names are the property of their respective owners and are used for identification purposes only.