# Share of Model

## Quick answer

Share of model is how often a brand is mentioned by one specific AI model, such as ChatGPT or Claude, compared to its named competitors, as opposed to an average across every AI engine combined. Models differ in what they cite and how they describe a brand, so an average can hide a serious gap on one of them.

## Why it matters

A brand could average a reasonable AI share of voice across several engines while actually being invisible on the one engine its specific buyers use most. Breaking the number down per model, rather than reporting one blended figure, is what catches that, since an average flattens exactly the differences that matter most for deciding where to focus limited time and attention, and a team that only looks at the average can miss a severe, fixable gap for months.

## How to measure it

Track the same buying questions separately for each AI model rather than blending the results into one combined score, then compare the brand's standing on each model against its named competitors on that same model, for the same question set, on the same schedule, so a gap on one specific model is visible rather than averaged away into a reassuring overall number.

## Example

A legal-tech tool's blended AI visibility score looks healthy at first glance. Broken out by model, it's strong on two engines and essentially absent on a third, specifically the one that happens to be most popular with its target buyers. The blended average made that gap invisible; the per-model breakdown didn't, and pointed the team at exactly where to focus next instead of spreading effort evenly across every engine regardless of where the actual problem sat.

## FAQ

### Why not just use one combined AI visibility score?

Because a combined score can look fine on average while hiding a severe gap on one specific model, especially if that model happens to be the one a brand's actual buyers use most for their own research.

### Do all AI models describe brands the same way?

No. Each model draws on different sources and training, which means the same brand can be described very differently, or not mentioned at all, from one model to the next, even when asked the exact same question.

### Which models should share of model be tracked across?

Whichever ones a brand's buyers actually use. MarketHQ tracks this across six AI engines: ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek.

### Does a weak share of model on one engine mean the brand failed there?

Not necessarily a failure, more often a gap in what that specific engine has read or indexed. The fix usually starts with checking what that engine actually cites for the same question, then addressing whatever page is missing or underperforming for it specifically.

## Related terms

- /glossary/share-of-voice
- /glossary/ai-share-of-voice
- /glossary/share-of-market

## Related

- LLM share of voice for SaaS: /blog/llm-share-of-voice-for-saas
- AI visibility use case: /use-cases/for-ai-visibility