AI shopping has its own taste, shortcuts, and failure modes. It compares the options before the shopper ever sees a storefront.
How AI shopping systems decide what gets recommended.
Abstract
Shopping is moving from the search bar into ChatGPT and Gemini. Caeliai measures how those systems choose products, where they send the buyer, and what ecommerce brands can change to win the recommendation. We publish the research, then help brands act on it.
§ 01 — The shift
The buying decision moved from the search page into the chat.
A shopper describes what they want in one sentence and gets back product cards with prices and buy buttons. That answer is the new shelf — and it is where the buying decision now happens.
Product data now shapes model-generated answers, not just feeds and search rankings.
Visibility is measurable. Most brands have never measured it.
§ 02 — Findings
AI names the brand. The sale still goes somewhere else.
GENESIS-03 followed real ChatGPT and Gemini shopping conversations across 11 ecommerce categories and scored where every recommendation actually sent the buyer.
GENESIS-03 · N = 380 brand observations · 11 categories
63%
of AI shopping recommendations don’t send the buyer to the brand’s own store.
Winner 37%
Leak 18%
Loser 45%
Winner
The buy path lands on the brand’s own product page. Owned demand.
Leak
The buy path routes to a retailer or marketplace. Someone else gets the sale.
Loser
The brand is named with no usable buy path at all. A recommendation that goes nowhere.
§ 03 — Published work
Research papers
Genesis 01: a vision benchmark for fashion AI.
Benchmarks CLIP, SigLIP, and DINOv2 on high-fashion nuance: uncertainty, impostor detection, and collection cohesion.
Genesis 02: what LLMs repeatedly choose.
Maps aesthetic consensus across nine language models — the designers, garments, and narratives they converge on from prompts alone.
Genesis 03: AI shopping visibility and the buy path problem.
Follows ChatGPT and Gemini recommendations into the routing layer: official wins, third-party leaks, and missing purchase paths.
Articles
How does AI shopping work for ecommerce brands?
How a shopper’s sentence becomes a shelf of product cards, why AI shoppers convert 4–5× higher, and why 63% of recommendations never reach the brand’s own store.
§ 04 — Applied
The research becomes advisory.
The same instruments we built for the Genesis studies, pointed at your brand: real shopping prompts, scored buy paths, fixes ranked by what actually moves recommendations.
01
Audit
See what AI systems currently say: the prompts you appear in, the competitors named beside you, and where every buy path lands.
02
Repair
Fix the product data and routing layer the models read — feeds, structured data, PDPs, and destination links.
03
Measure
Track recommendation share over time and tie AI-originated discovery back to traffic and revenue.
Free
Where is ChatGPT sending your buyers?
We run real shopping prompts for your category and show you where the recommendations land — your store, a reseller, or nowhere. One form, no meeting.