> ## Documentation Index
> Fetch the complete documentation index at: https://docs.shareofmodel.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Pencil Agent — User Guide

> Use the Share Of Model agent inside Pencil to analyse brand perception and guide content creation.

The **Share Of Model Agent** inside Pencil is a specialised agent that analyses brand perception across LLMs and surfaces actionable insights for improving brand positioning and content.

## What the agent does

<CardGroup cols={3}>
  <Card title="Brand perception insights" icon="eye">
    Analyse how your brand is perceived across LLMs (Claude, Gemini, GPT, Deepseek, and others).
  </Card>

  <Card title="Improvement recommendations" icon="wand-magic-sparkles">
    Suggest how to improve brand perception based on the data.
  </Card>

  <Card title="Content optimisation" icon="lightbulb">
    Guide how to optimise images, PDPs and other content for specific LLMs.
  </Card>
</CardGroup>

Once you have insights, hand off to specialised content-generation agents (Image Generation, Video Generation, etc.) to produce the recommended assets.

## Data sources

The agent draws on Share Of Model analyses run for the brand. For each analysis, it has access to:

### Brand perception data

* **Brand perception attributes** — each attribute represents a positive (strength) or negative (weakness) brand perception.
* **Brand perception cluster groups** — thematic groupings of similar attributes.
* **Brand perception by source** — strengths and weaknesses broken down by LLM (Claude, Gemini, GPT, etc.).
* **Brand perception by brand** — comparative view of your brand vs. competitors.

### Target audience data

* **Audience segments** — demographics, behavioural attributes, and media consumption.
* **Age range distribution** — statistical distribution across age ranges.

### Analysis metadata

* Brand under analysis, geographic countries, category, competitors, personas, and LLM sources (Anthropic, Google, OpenAI, Deepseek, etc.).

## Getting started

### Prerequisites

To start using the agent, you need to:

<Steps>
  <Step title="Select a brand">
    The agent lists all available brands. If only one brand is available, it is selected automatically.
  </Step>

  <Step title="Select an analysis">
    Pick a Share Of Model analysis to work with. If only one exists, it is selected automatically.
  </Step>
</Steps>

### Initial setup workflow

When you first interact with the agent:

1. The agent automatically lists available brands.
2. If multiple brands exist, you are prompted to select one.
3. The agent then lists available analyses for the selected brand.
4. If multiple analyses exist, you are prompted to select one.
5. Once selected, the analysis is stored in the conversation context.

## Key capabilities

<Tabs>
  <Tab title="Brand perception">
    Analyse perception across multiple dimensions:

    * **Overall cluster share** — distribution of perception clusters (strengths and weaknesses).
    * **Cluster share by brand** — your perception clusters vs. competitors.
    * **Cluster share by source** — how different LLMs perceive you.
    * **Attributes hierarchy** — detailed hierarchy of attributes within clusters.

    **Example queries**

    * *What are the main strengths and weaknesses of my brand?*
    * *How does my brand compare to competitors in terms of perception?*
    * *Which LLM sources have the most positive perception of my brand?*
  </Tab>

  <Tab title="Source-specific">
    Analyse how your brand performs on specific LLM platforms:

    * identify the most favourable sources,
    * understand source-specific strengths and weaknesses,
    * get recommendations for optimising content per platform.

    **Example queries**

    * *What are my brand's strengths on Claude vs. Gemini?*
    * *How should I optimise my content for GPT?*
  </Tab>

  <Tab title="Audience analysis">
    Understand your target audience segments:

    * demographic breakdowns,
    * behavioural patterns and preferences,
    * media consumption habits,
    * age distribution.

    **Example queries**

    * *Who is my target audience?*
    * *What are the main challenges of my audience segments?*
    * *What media channels does my audience prefer?*
  </Tab>

  <Tab title="Content recommendations">
    Get actionable content guidance based on the analysis:

    * attributes to emphasise,
    * content themes that align with strengths,
    * areas to address to fix weaknesses,
    * LLM-specific optimisation strategies.

    **Example queries**

    * *How can I improve my brand perception through content?*
    * *What should my next image campaign focus on?*
    * *What messaging should I use for my PDP?*
  </Tab>
</Tabs>

## Integration with content-generation agents

The agent is designed to feed into Pencil's content-generation agents.

### Typical workflow

<Steps>
  <Step title="Analysis (Share Of Model agent)">
    Analyse perception, identify strengths and weaknesses, get content recommendations.
  </Step>

  <Step title="Generation (specialised agents)">
    Use the insights to create image campaigns, generate videos aligned with brand strengths, build optimised PDPs, or draft email copy.
  </Step>
</Steps>

### Examples

<AccordionGroup>
  <Accordion title="Creating new assets">
    **User:** *I want to create an image campaign to improve my brand perception.*

    The Share Of Model agent identifies key strengths (e.g. *innovative*, *sustainable*, *premium quality*) and weaknesses (e.g. *price perception*, *accessibility*), then recommends focusing visuals on innovation and sustainability.

    The user switches to the Image Generation agent and uses those insights to generate images that emphasise innovation and sustainability while addressing accessibility concerns.
  </Accordion>

  <Accordion title="Improving existing assets">
    **User:** *I have an existing product image. How can I improve it based on brand perception?*

    The Share Of Model agent reviews current perception across LLMs, finds that *premium quality* is a strength under-represented and *accessibility* is a weakness, and recommends enhancing visual cues to emphasise premium quality while making the product appear more accessible.

    The user opens the Image Generation agent with the existing image as input. The agent edits or regenerates the image with refined lighting, premium textures, and clearer, more approachable styling — better aligned with brand perception goals.
  </Accordion>
</AccordionGroup>

## Best practices

1. **Start broad** — begin with overall brand perception before diving into specific dimensions.
2. **Compare across sources** — surface platform-specific opportunities by comparing LLMs.
3. **Use competitive context** — leverage competitor comparisons to understand relative positioning.
4. **Link insights to content** — always connect insights to actionable content recommendations.
5. **Iterate** — create content from recommendations, measure impact, then refine.

## Response format

The agent returns:

* **Executive summary** — high-level findings.
* **Detailed analysis** — data breakdowns and metrics.
* **Actionable recommendations** — clear next steps for content creation.
* **Context** — data sources, time periods, and limitations.

All shares and percentages are expressed as percentages for clarity.

## Language support

The agent can translate insights and recommendations into your preferred language — request translation in the conversation.

## Error handling

If data is unavailable or incomplete, the agent will:

* transparently report limitations,
* suggest alternative approaches,
* continue analysis with available data while noting the gaps,
* estimate the impact of missing data on conclusions.

## Terminology

| Term               | Definition                                                    |
| ------------------ | ------------------------------------------------------------- |
| **Strengths**      | Positive brand perception attributes (preferred over "pros"). |
| **Weaknesses**     | Negative brand perception attributes (preferred over "cons"). |
| **Cluster groups** | Thematic groupings of similar attributes.                     |
| **Source**         | LLM provider (e.g. Claude, Gemini, GPT).                      |
| **Analysis**       | A Share Of Model analysis containing brand-perception data.   |
| **Persona**        | Target audience persona defined in the analysis.              |

## Example use cases

<AccordionGroup>
  <Accordion title="Brand perception audit">
    **User:** *Analyse my brand's perception across all LLM sources.*

    The agent lists strengths and weaknesses, compares perception across sources, provides percentage breakdowns, and recommends focus areas for improvement.
  </Accordion>

  <Accordion title="Competitive analysis">
    **User:** *How does my brand compare to competitors?*

    The agent provides a cluster-share comparison by brand, identifies competitive advantages, highlights areas where competitors outperform, and suggests differentiation strategies.
  </Accordion>

  <Accordion title="Content strategy">
    **User:** *What content should I create to improve perception on Gemini?*

    The agent analyses Gemini-specific perception, identifies the attributes to emphasise, provides content themes and messaging recommendations, and suggests switching to the Image or Video Generation agent for execution.
  </Accordion>

  <Accordion title="Audience-targeted content">
    **User:** *What content resonates with my target audience?*

    The agent analyses audience segments, identifies key challenges and preferences, recommends content themes aligned with audience needs, and provides media-channel recommendations.
  </Accordion>
</AccordionGroup>

## Support

If you encounter issues:

* **No analysis available** — ensure a Share Of Model analysis has been created and is accessible.
* **Missing data** — the agent reports limitations and suggests alternatives.
* **API errors** — check the [API Reference](/api-reference/introduction) for status and requirements.

## What's next

<CardGroup cols={2}>
  <Card title="Pencil Pro Scoring" icon="pen-ruler" href="/platform/evaluations/share-of-model-scoring-in-pencil-pro">
    Score creatives directly inside Pencil.
  </Card>

  <Card title="API Reference" icon="code" href="/api-reference/introduction">
    Drive the same data programmatically.
  </Card>
</CardGroup>