Early access
Start free trial
Glossary · Metric

What is share of model?

The primitive metric of AI visibility — how often a single engine surfaces a brand or product across a defined prompt set.

Last updated June 2026

In detail

The arithmetic is deliberately simple: a brand with 20 mentions across 100 ChatGPT responses has a 20% share of model on ChatGPT for that prompt set. Run the identical prompts on Perplexity and the number will differ — the engines retrieve from different source pools and weight them differently, so the metric is only meaningful with the engine attached.

Share of model is a primitive, in the database sense: one raw number per engine that every higher-level report is built on. Competitive share of voice, category leadership, week-over-week deltas — all derive from it. It adapts the classical share-of-voice idea to a surface where the denominator is generated answers rather than ad impressions.

Because generative answers vary run to run, the number is only as trustworthy as its sampling: a fixed prompt set, a consistent cadence, and enough runs to smooth variance. One-off spot checks produce a share-of-model anecdote, not a metric.

Why it matters for ecommerce

Per-engine resolution is the entire point. A brand at 15% share of model on ChatGPT and 5% on Perplexity has a Perplexity problem that a blended average would have buried — and the fix differs by engine, because PDP copy that moves ChatGPT does not reliably move Perplexity, which leans harder on third-party citation tails.

For a VP of Ecommerce, share of model on buying-intent prompts is the leading indicator that sits closest to revenue: it predicts AI-referred traffic and agent-drafted carts before either shows up in analytics. For the SEO lead, it is the AI analog of an impression share report — familiar enough to defend in a monthly review.

Example

A hair-serum brand tracks a 40-prompt set on ChatGPT and Perplexity. Week one: the flagship 30ml bottle appears in 7 of 40 ChatGPT responses (17.5%) and 14 of 40 Perplexity responses (35%). The gap is the diagnosis — Perplexity finds the PDP easily, while ChatGPT's answers favor competitors with stronger review-site coverage. That week's work list becomes review-placement outreach, not another PDP rewrite. Illustrative numbers; the read pattern is the point.

How eCommerce Insights computes it

Share of model is computed per SKU, per engine, per week (daily on the Growth plan), with the full answer history retained from prompt tracking runs. SKU numbers roll up to brand level on demand, and the per-product reading feeds each SKU's citation score. The dashboard view is described under SKU-level tracking.

Related terms

Ask AI about share of model

Have your favorite AI engine apply this definition to your catalog.

Frequently asked questions

How is share of model calculated?
Mentions divided by total responses, expressed as a percentage. Run a fixed prompt set against a single engine, count the responses that name the brand or SKU, and divide by the total. Most dashboards roll the number up per week and per prompt cluster so teams can see which topics are pulling visibility up or down.
Why report per model instead of blending engines?
Each engine retrieves and ranks sources differently. A SKU can score 40% on Perplexity and 5% on ChatGPT for the same prompts; a blended number hides exactly where the work is needed. Per-engine reporting lets a team triage ChatGPT-specific fixes — usually third-party review coverage — separately from Perplexity-specific ones.
What is a good share of model number?
No published industry benchmark exists as of mid-2026. The usable comparisons are internal trend (this SKU versus last month) and the competitive read alongside share of voice. As a rough orientation, a flagship SKU below 10% on category-typical buying prompts usually has fixable PDP gaps; above 30% it is defending, not building.
Should share of model be measured per brand or per SKU?
Both, but the per-SKU reading is the one that maps to revenue. Brand-level share of model says the company appeared; SKU-level share of model says which product appeared — which is what merchandising, margin math, and the fix queue actually need. eCommerce Insights computes it per SKU per engine and rolls up to brand on demand.
How is share of model different from share of voice?
Share of model is absolute: mentions out of total responses on one engine. Share of voice (AI) is relative: the brand's slice of all mentions across a named competitor set. Share of model says you appear in 18% of ChatGPT answers; share of voice says you hold 18% of the mentions that went to your five-brand competitive set.

Go deeper

See share of model per SKU, per engine, per week. Start the 14-day free trial — no credit card.