Share of Model shows you how your brand appears in the answers people get from AI models such as ChatGPT, Gemini and Perplexity. You can request access through your Customer Success Manager. There is no extra subscription cost: Share of Model draws on the generations already included in your Pencil subscription, and there is no separate credit type to manage.

Share of Model runs as its own tool, on a separate URL from Pencil, part of the same suite and billed through your Pencil generations. The visibility gaps it finds feed directly into Pencil: once you spot where your brand is missing from AI answers, you can build and test the assets to close that gap without switching tools.

This article explains how each part of Share of Model counts generations, so you can plan your usage with confidence.

The Share of Model Agent allows you to run custom Share of Model analyses and queries inside Pencil. Each run of the agent counts as two generations.

For how agents and scores consume generations more broadly, see: How Are Generations Counted?

Share of Model connects your creative library in Pencil with the Share of Model platform, so you can request Asset Evaluation Scores for any of your creatives without leaving your workspace.

Each score assesses how well a creative reflects your brand's strategic attributes, for example Innovation or Trust, based on how large language models perceive it.

It helps you validate brand alignment before launch, spot brand cues that are missing or overrepresented, prioritise creative versions for testing, and support stakeholder feedback with objective, LLM-based evaluations.

How it works:

Getting set up: reach out to your Customer Success Manager, who activates Share of Model for your workspace and completes the setup with you.

When you use Share of Model directly, consumption depends on the module you run and the model it uses. The most important distinction is how often each module counts generations.

Weekly modules: Search, Brand and Shopping monitor your brand on a recurring schedule. They run every week and count generations every week for as long as monitoring stays active.

One-off modules: GEO audit and Assets count generations once for each item, with no recurring charge.

In Search, you define a set of prompts, the questions you want tested, and choose which AI engines to run them on. Search runs the latest version of each engine.

Each engine has a weekly prompt allowance that counts as one generation. The allowance counts as one generation whether you use all of it or only part of it.

A brand analysis runs weekly for each persona or country you track. The number of generations depends on the model running the analysis.

Because the analysis repeats weekly, the count scales with what you track. For example, tracking three personas across two countries on Claude 4.6 Sonnet runs six analyses and counts twelve generations each week.

The model also changes the count, and you can track more than one. Running those same six analyses on Gemini 3 Flash counts six generations each week, and running them on both models together would cost eighteen generations .

Shopping monitors how your brand appears in Rufus, Amazon's shopping assistant. One generation per week covers up to 50 prompts.

A GEO audit analyses a URL to assess how well it performs as a source for AI answers. Each URL analysed counts as one generation, covering up to 100 prompts for that URL. This is a one-off charge per URL, with no recurring cost.

The Assets module counts generations once for each item generated. The rate depends on the asset type and the model.

Share of Model consumption comes out of the same generation balance as everything else in Pencil. You can check your remaining generations with the counter in the bottom right corner of your workspace. For details, see: How Are Generations Counted?