Methodology
AI Commerce glossary & methodology.
The terms, measures and principles behind Share of Conversation — explained for commercial teams.
AI assistants, AI search and retailers’ own AI are changing how shoppers discover products. Traditional digital shelf metrics were not built to see inside those answers.
This page explains how Share of Conversation measures the brands Direct AI puts forward, which AI environments we look at, and the principles that keep the evidence reliable.
Glossary
Key terms.
Core KPI · Direct AI
Definition
Gemini
OpenAI
Perplexity
Why it matters
It shows who AI is putting forward in your category — as a competitive view, read against every other brand, not an isolated score.
In practice
Measured across Direct AI — Gemini, OpenAI and Perplexity — and shown as a percentage. Every brand is measured against the same category evidence base.
Supporting terms
Definition
The set of real shopper questions a benchmark is built on, covering the needs and missions that matter in the category.
Why it matters
Measuring a robust, representative question set — rather than a handful of ad-hoc prompts — is what makes the result a benchmark.
Definition
A mention of a brand in an AI answer that we can confidently match to a brand in the category’s competitive field.
Example
Real benchmark · Sun Protection & SPF UK
In our Sun Protection & SPF UK benchmark, 480 Direct AI answers produced 2,087 counted brand references across 172 brands in the measured field.
Why it matters
Only references we can recognise with confidence are counted. Uncertain evidence is handled conservatively.
Definition
The set of brands AI puts forward for a category, and how Share of Conversation is distributed between them.
Why it matters
The field can change by AI provider and by shopper need — which is where competitive differences become visible.
Environments
AI Commerce environments.
One shopper need can pass through several AI decision environments. Share of Conversation is measured in Direct AI; the AI Commerce Benchmark treats every other environment as a separate evidence lens, measured where evidence is available.
Separate evidence lens
Shopper demand
Evidence of what shoppers are looking for and how they ask.
What we measure
separate demand-intelligence evidence, where measured. It is reported on its own terms and is not part of Share of Conversation.
Share of Conversation measured here
Direct AI
Gemini
OpenAI
Perplexity
AI assistants that answer shopper questions directly.
What we measure
Share of Conversation across Gemini, OpenAI and Perplexity — the core of every benchmark.
Separate evidence lens
AI Search
AI-generated answers within search, such as Google AI Mode.
What we measure
AI Search evidence, where measured, reported separately from Direct AI Share of Conversation.
Separate evidence lens
Retailer AI
Retailers’ own shopping assistants — such as Amazon Rufus or Alexa for Shopping, depending on the market, and Boots AI Assistant.
What we measure
Retailer AI evidence, where specifically measured, reported separately.
Separate evidence lens
Product / digital shelf
The product pages and digital shelf a shopper lands on.
What we measure
diagnostic evidence, where measured, reported separately. It sits alongside Share of Conversation and does not explain it.
Share of Conversation · Direct AI
Separate evidence lens, where measured
Lenses
Reading the evidence.
Share of Conversation
Read by provider
Illustrative
Provider view
What it shows
Share of Conversation read separately for Gemini, OpenAI and Perplexity. It shows how the competitive field changes from one AI to the next — without ranking the providers or combining them into one score.
Read by shopper need
Illustrative
Shopper-need view
Question group
Question group
Question group
What it shows
Share of Conversation read across groups of related shopper questions, showing where the field shifts as needs change.
Where measured
Alongside the answers · observation, not proof
Sources and themes
The sources and themes that appear alongside AI answers. We treat these as observations worth investigating — not as proof of why a brand was put forward.
Where measured
Illustrative
Boundaries
What Share of Conversation is not.
Share of Conversation is a share of real brand references in a defined category. It is not a 0–100 score, a composite or a blended AI visibility index.
It doesn’t measure search rankings, advertising share or SEO performance. It measures which brands Direct AI puts forward when shoppers ask.
Principles
Our principles.
01
Built on real shopper questions
Every benchmark starts with a certified set of questions that reflect how shoppers actually ask about the category.
02
Measured consistently
Each question is put to each Direct AI provider in the same way. Brand names are standardised so every reference is counted against the right brand.
03
Conservative by design
Uncertain evidence is handled cautiously, and missing evidence is never treated as zero. There is no hidden weighting.
04
A point-in-time view
AI answers change. Each benchmark reports what was observed when it was measured, and states its scope — which environments and providers were included.