Share of model: the metric replacing rank tracking

Share of model is the percentage of relevant AI answers in which your brand is mentioned or recommended, measured across a fixed set of buyer prompts and a fixed set of assistants. If you run 200 prompts across five assistants each month and your brand appears in 410 of the 1,000 answers, your share of model is 41 percent.

For twenty years, position in the search results was the unit of SEO reporting. It still matters, but it no longer describes the whole picture. A buyer who asks ChatGPT for the best option in your category never sees a ranking. They see three or four names. Share of model tells you how often one of them is yours.

Why rankings alone stopped working

• The answer moved above the results. AI Overviews and AI Mode answer many queries before a searcher reaches a link.
• Research moved into assistants. A meaningful share of product research now happens in ChatGPT, Perplexity and Copilot, where there are no positions.
• Click-through decoupled from position. A stable number-one ranking can lose half its clicks when an Overview appears above it.

None of this shows up in a rank tracker. Share of model is the metric that does.

How to define it for your business

Choose the prompt set

Draw prompts from real buyer language: sales call notes, support tickets, community threads and search data. Cover the funnel: category discovery ("what tools do X"), comparison ("X versus Y"), recommendation ("best X for Y") and validation ("is X good for Y"). Aim for fifty to two hundred prompts and freeze the set so months are comparable.

Choose the assistants

ChatGPT, Perplexity, Claude, Gemini and Copilot cover most usage. Include Google AI Overviews if you want one number across search and assistants.

Decide what counts

Record four things per answer: whether you were mentioned, your position if a list was given, whether the description was accurate, and the sentiment. A mention that says you are expensive and outdated is not the same as a recommendation.

The calculation

Metric	Formula	Use
Share of model	Answers mentioning you ÷ total answers	Headline visibility
Recommendation rate	Answers recommending you ÷ answers with a recommendation	Quality of presence
Average position	Mean list position when named	Shortlist strength
Accuracy rate	Accurate descriptions ÷ mentions	Brand control
Citation share	Answers citing your pages ÷ total answers	Content performance

Why it works as a board metric

• It is one number, comparable month to month and against competitors.
• It reflects the moment closest to a buying decision.
• It moves in response to work you control: content, authority and third-party presence.

How to run the measurement

1. Run every prompt in every assistant on a fixed schedule with consistent settings.
2. Capture the full answer and cited sources.
3. Score mentions, position, accuracy and sentiment with a consistent rubric, using automation for the first pass and a human check on a sample.
4. Report the trend, not a single run. Individual answers vary; the monthly aggregate is stable.

Our free AI Visibility Checker runs a five-prompt version of this so you can see the format before committing to a full programme.

Reading the numbers

A few patterns come up repeatedly once teams start tracking.

• High share, low recommendation rate. You are known but not preferred. The fix is usually comparison content and reviews that state why you win for specific use cases.
• Presence in one assistant but not others. Retrieval differs by engine. Perplexity leans heavily on live search results; ChatGPT leans more on training knowledge. Check which sources each one cites and fill the gap that engine relies on.
• Accuracy below 80 percent. Stale third-party pages are describing you. Correct them at the source and publish a clear, current description on your own site.
• Sudden drops after a model update. Expected. Judge the trend over three months, not one.

Segment the prompt set by funnel stage. A brand can dominate validation prompts ("is X good for Y") while being absent from discovery prompts ("what tools do Y"), and the two need different work.

Pairing it with revenue

Share of model is a leading indicator. Pair it with organic pipeline and revenue so the board sees both the cause and the effect. When share of model rises in a category and assisted pipeline follows a quarter later, the case for continued investment makes itself. This is the reporting model we use in every answer engine optimization engagement.

To understand the work that moves the metric, read how to get your brand recommended by ChatGPT.

Frequently asked questions

What is a good share of model?

It depends on category size. In a category with five serious competitors, above 40 percent usually means you are on most shortlists. The more useful benchmark is your trend and your gap to the leader.

How many prompts do I need?

Fifty to two hundred is typical. Fewer than twenty produces noisy month-to-month swings; more than a few hundred adds cost without much extra signal.

Which assistants should I track?

ChatGPT, Perplexity, Claude, Gemini and Microsoft Copilot cover most buyer usage. Add Google AI Mode and AI Overviews if you want a single view of search and assistants.

Is share of model the same as share of voice?

It is the AI-answer equivalent. Share of voice measures presence in search results or media; share of model measures presence in AI-generated answers.