Share of Model measures how much of AI assistants' brand recommendation space belongs to your brand versus its competitors. It is the practical equivalent of share of voice, applied to answers generated by ChatGPT, Gemini and other assistants.
What Share of Model measures
It starts with a stable set of commercial questions, such as “best invoicing software for freelancers”, “alternatives to X” or “which tool should I choose for Y”. Each answer is checked for brands and position.
The basic formula is:
Your brand mentions / total mentions of all brands × 100
If your brand appears 18 times and the competitive set totals 90 mentions, your Share of Model is 20%. It can be calculated globally or by assistant, country, category and search intent.
Share of Model is not mention rate
- Mention rate: the percentage of tracked questions where your brand appears.
- Share of Model: your brand's weight among all recommended brands.
- Average position: where you appear inside answers that mention you.
The three work together. A brand may appear frequently but always last; another may appear less often while leading the highest-intent questions.
How to build a reliable measurement
- Define 15 to 40 questions across discovery, comparison, pricing, objections and purchase.
- Keep their wording identical over time.
- Query the same assistants and record the date.
- Count mentions, position and competitors in every answer.
- Repeat weekly or monthly and compare trends, not isolated screenshots.
Do not add new questions halfway through a series without separating them: changing the universe changes the percentage even if brand presence stays flat.
How to read the result
There is no universal grade for an isolated percentage. Twelve percent may be strong in a twenty-player category and weak in a three-player market. Use three references: your previous result, the distance to the leader and performance on commercially valuable questions.
It is also useful to separate ChatGPT and Gemini. Assistants may rely on different sources and produce different rankings; a global average can hide a specific opportunity.
How to increase it
Start with questions where a competitor appears and you do not. Identify the cited sources, the reason that brand is recommended and the page on your domain that should answer the intent better.
Common actions include improving an existing page, clarifying the brand entity and category, publishing verifiable comparisons, strengthening structured data or earning presence in relevant third-party sources.
Turning it into a useful series with AskBrandly
AskBrandly checks mentions across a consistent question set, identifies competitors ahead of you and surfaces cited sources. The value is not asking once; it is maintaining a comparable benchmark to see whether completed work earns real share.
For the wider scorecard, read how to analyse competitors in AI answers, how to measure brand perception and how to audit your domain's AI visibility.