# Guided prompts

Where [tools](/docs/mcp) are low-level primitives, **prompts** are named, task-oriented starting points. They appear in your assistant's prompt or slash menu with a description and argument hints, and each one tells the assistant exactly which tools to chain to complete a full task. They're the recommended way to begin — pick one, fill in its argument, and let the assistant do the orchestration.

<Callout>
Prompts don't do anything you couldn't ask for in plain language — they're just reliable, pre-written recipes. If a prompt doesn't quite fit, ask your question directly and the assistant will select tools itself.
</Callout>

## Report prompts

These take a report's **public token** (from a share URL) and analyze a single shared report. See [Reports](/docs/mcp/reports) for the underlying tools.

| Prompt | Argument | What it does |
|---|---|---|
| **Analyze a Spyglasses report** (`analyze_report`) | `publicToken` | Detects whether the token is an AI Visibility report or a Site Readiness audit, then leads with the headline numbers and the 3–5 most important findings. |
| **What should I fix first?** (`what_to_fix_first`) | `publicToken` | Turns a report into a prioritized action plan grouped by team (SEO/AEO, PR, technical), highest-impact and lowest-effort first. |
| **Where do competitors beat me?** (`competitor_gaps`) | `publicToken` | Surfaces the searches and citations where competitors appear and the brand doesn't, with content and PR targets to close the gaps. |
| **Find my weakest pages** (`audit_worst_pages`) | `publicToken` | Ranks a site audit's lowest-scoring pages and explains the specific fixes to raise each one's readiness score. |

## Account prompts

These use your signed-in account. Most take a `propertyId` or `projectId` you get from `list_properties` / `list_projects`. See [Account data](/docs/mcp/account-data) and [Scoring](/docs/mcp/scoring) for the underlying tools.

| Prompt | Argument | What it does |
|---|---|---|
| **List my organization's reports** (`list_my_reports`) | — | Lists the reports in your organization (asking which one if you belong to several) so you can pick one to analyze. |
| **Analyze a project** (`analyze_project`) | `projectId` | Summarizes a project's progress — metric deltas, goals, weekly trend — and recommends next moves. |
| **Track message drift** (`track_message_drift`) | `propertyId` | Shows how AI's description of the brand changes over time for a tracked query, plus how key-message pull-through trends. |
| **Citation mix over time** (`citation_mix_trends`) | `propertyId` | Shows how the mix of sources AI cites shifts over time, and where to focus content and outreach. |
| **Evaluate publishers (AIPVS)** (`evaluate_publishers`) | `domains`, optional `propertyId` | Scores and ranks publisher domains by AI Placement Value Score, optionally in a brand's context. |

## How prompts relate to tools

Each prompt expands into a short plan the assistant follows. For example, **Track message drift** runs roughly:

<Steps>

<Step>
Call `get_message_tracking` to see each key message's pull-through rate over time.
</Step>

<Step>
Call `get_answer_summaries` with no query to list tracked queries, pick the most strategic one, then call it again for that query's full weekly answers per platform.
</Step>

<Step>
Compare the weekly answers to narrate what's changing, and chart pull-through from the returned series.
</Step>

</Steps>

You can always stop a prompt partway and steer — ask for a different query, a chart, or a deeper drill-down. The prompt just gets the conversation moving in the right direction.

## Related

- [Reports](/docs/mcp/reports) — the token-scoped report tools the report prompts use
- [Account data](/docs/mcp/account-data) — the property tools the account prompts use
- [Scoring](/docs/mcp/scoring) — publisher and placement scoring behind `evaluate_publishers`