GUEST COMMENT Retailers are about to find out whether they still own their own shelves

Topic
Retail
Published

This summary was written by ShopDigest AI from InternetRetailing’s report, by Paul Skeldon. Read the full story ↗

James Taylor explores the real issue at the heart of the “AI in retail media” debate: who controls the decision at ranking time? When a shopper asks an AI assistant for “moisturiser for sensitive skin under £30”, the first shelf they see may no longer be assembled on a retailer’s website. It may be assembled […]

Overview

The article argues that as AI shopping assistants become the primary discovery interface, retailers risk losing control over which products are shown to shoppers. The author contends that the real battleground is the ranking decision, which must remain under retailer control to protect margins and commercial priorities. It distinguishes five AI functions and advocates for a retailer-governed decisioning layer between conversational AI and the product catalogue.

Key Details

  • Five AI functions: causal measurement, predictive ML, decisioning, generative AI, and agentic orchestration — each with distinct roles.
  • Raw catalogue exposure to an LLM cedes ranking control to external systems, risking profit loss.
  • A retailer-controlled ranking layer can integrate relevance, stock, margin, supplier funding, and merchandising rules.
  • Case evidence: a pet retailer saw ~4x monetisation coverage with transformer-led relevance; a beauty retailer saw 9x impressions but 87.3% lower CTR and 55.6% lower sales with bid-only auctions.
  • AI agents will abandon retailer endpoints that produce irrelevant results, making trust in ranking quality essential.

Why It Matters

For merchants, the shift to AI-driven discovery means the shelf is no longer a webpage but a ranked list inside an assistant. Retailers must invest in decisioning infrastructure that keeps commercial control in-house, or risk losing margin and relevance to external ranking systems. The article urges optimising for retailer profit, not proxy metrics like impressions or clicks.

Frequently asked questions

What is the main concern about AI in retail media according to the article?

The main concern is that retailers may lose control over which products are shown to shoppers when AI assistants assemble the first shelf. The article argues that whoever controls the ranking decision holds commercial power, and retailers must own that layer to protect margins and influence.

What are the five AI functions the article distinguishes?

The five functions are causal measurement (determining if a campaign caused an outcome), predictive ML (estimating future probabilities), decisioning (combining signals to decide what to show), generative AI (creating language and imagery), and agentic orchestration (coordinating multi-step workflows). The article stresses that decisioning is where monetization happens and should not be conflated with the others.

Why does the article say exposing a raw catalogue to an LLM is risky?

Exposing a raw catalogue to an LLM without a control layer turns the catalogue into input for someone else's ranking system, potentially an external auction. This can lead to loss of profit potential because the retailer no longer governs which products are eligible, relevant, or commercially prioritized.

What evidence does the article provide about the impact of AI-driven ranking?

The article cites two examples: a pet retailer saw monetization coverage increase about four times after switching to transformer-led relevance ranking, and a beauty retailer saw ad impressions rise nine times but click-through rate fall 87.3% and sitewide sales drop 55.6% when using bid-only auctions. These illustrate that relevance-driven ranking can improve outcomes while bid-only auctions can degrade value.

What does the article suggest retailers should do to maintain control?

Retailers should implement a retailer-controlled ranking layer that evaluates shopper intent alongside relevance controls, stock, margin, supplier funding, campaign pacing, and merchandising rules. This layer should integrate paid and organic priorities into one engine, ensuring commercial influence is proportional to relevance and paid placements are clearly identified.

About this story

Published
17 Aug 2026
Originally published at