MCP tool referenceReads and changes
ai_visibility_audit: how well AI search knows your brand
ai_visibility_audit reads and runs a brand's AI Visibility audit: the questions it tracks, how AI engines answer them, the Visibility Score, mention and citation rates, and the ranked list of next actions. A brand has at most one audit.
Read it with a view. Change it with an action: set the audit up, change its settings, add, remove or run prompts, and tick off next actions.
Views and actions01
What it can read and do
Every call names a brand and either a view (what to read) or, on the two audit tools, an action (what to change). With neither, the tool returns its default view.
Views (reads)
reportDefault- The Visibility Score, mention and citation rates, completeness, accuracy, topics, cited sites, the strategy plan and clusters. Narrow it with cluster, topic, objective, funnel_stage or user_intent.
prompts- The tracked questions with their status and last result. Filter by status, cluster or text, and page through them.
clusters- The audience clusters, optionally for one AI engine.
actions- The ranked next-action checklist, reconciled with the latest results the way the dashboard does it.
status- Generation and run progress, to follow a setup or a run.
Actions (changes)
setupOwn cost- Create the audit. It writes prompt_count questions across audience clusters and runs each on every chosen engine. Costs prompt_count × engines credits, checked before anything starts, and takes a few minutes.
update_settings- Rename the audit, pause or resume it, or change the market, language, engines, and the objectives, funnel stages and user intents new questions are written for.
add_promptsOwn cost- Add your own questions. Each runs right away on every engine, at 1 credit per question per engine. Questions the audit already has are skipped.
remove_prompts- Remove questions by their exact text.
run_promptsOwn cost- Run or re-run a list of prompts, a whole cluster, or all_pending, with run_count 1 or 3. Costs prompts × engines × run_count credits.
update_action_status- Mark the next action titled item as done, dismissed or in progress, or reopen it. Pass cluster when two actions share a title.
Credits02
What a call costs
- Reads, with any view, cost 1 credit per call.
- update_settings, remove_prompts and update_action_status also cost 1 credit per call.
- setup, add_prompts and run_prompts skip that fee and charge what they cost in the dashboard: questions × engines, times run_count for run_prompts. setup checks the balance before it starts.
- Which engines you can use depends on the plan: Gemini on every plan, DeepSeek on paid plans, Claude on Enterprise.
Parameters03
Input parameters
Every parameter the tool's input schema accepts. Your assistant fills these in from what you ask; you never type them yourself.
| Parameter | What it does |
|---|---|
brandRequired | The brand's name, case-insensitive, as brand_insights lists it. |
view | What to read: report (default), prompts, clusters, actions or status. |
action | What to change: setup, update_settings, add_prompts, remove_prompts, run_prompts or update_action_status. Leave it out to read. |
cluster | An audience cluster name. Narrows a read; add_prompts files new questions under it; run_prompts runs every question in it; update_action_status uses it to tell same-titled actions apart. |
topic | view report: narrow to one topic. |
objective | view report: one growth objective, brand_authority, search_visibility or knowledge_optimization. |
funnel_stage | view report: one funnel stage, discovery, consideration or decision. add_prompts: tag the new questions with it. |
user_intent | view report: one user intent, informational, navigational, commercial or transactional. add_prompts: tag the new questions with it. |
engine | view clusters: one AI engine. |
status | view prompts: a prompt status (idle, pending, queued, running, completed, failed or timedout). update_action_status: the new status (done, dismissed, in_progress or reopen). |
search | view prompts: only questions containing this text. |
page | view prompts: the page number. |
limit | view prompts: rows per page. |
item | update_action_status: the action's title. |
prompts | add_prompts, remove_prompts, run_prompts: the questions, written the way a person would ask AI. |
all_pending | run_prompts: run every question that hasn't run yet. |
run_count | run_prompts: 1 or 3 runs per engine. Three runs average out how much AI answers vary. |
name | setup: the audit's name. Defaults to the brand name. |
new_name | update_settings: a new name for the audit. |
website | setup: the site to track citations for. Defaults to the brand's website. |
description | setup, update_settings: what the brand is and does. |
active | update_settings: false pauses the audit, true resumes it. |
market | setup, update_settings: where the audience is: indonesia, malaysia, vietnam or global. |
language | setup, update_settings: the language of the questions: id, ms, vi or en. |
engines | setup, update_settings: the AI engines to run on. Gemini is always available, DeepSeek needs a paid plan and Claude needs Enterprise. ChatGPT and Perplexity are coming soon and are refused for now. |
prompt_count | setup: how many questions to write and run. |
cluster_count | setup: how many audience clusters to build first. |
max_topics | setup: topics per cluster. |
keywords_per_topic | setup: keywords per topic. |
objectives | update_settings: the objectives new questions are written for. |
funnel_stages | update_settings: the funnel stages new questions are written for. |
user_intents | update_settings: the user intents new questions are written for. |
Example prompts04
Questions that use this tool
Ask in plain language and the assistant picks the view or action. These make good first questions.
- “What's [brand]'s Visibility Score, and what changed since the last run?”
- “Which tracked questions does AI answer without naming [brand]?”
- “Set up an AI Visibility audit for [brand] in Vietnam, in Vietnamese, with 30 questions on Gemini.”
- “Add these three questions to [brand]'s audit and run them.”
- “What are the top next actions for [brand]? Mark the first one done.”
FAQ05
Questions about this tool
How many audits can a brand have?
One. setup refuses if the brand already has an audit; change the existing one with update_settings.
How long does setup take?
A few minutes: it builds the clusters, writes the questions and runs them on every engine. Follow it with view status.
Why run a prompt three times?
AI answers vary from one run to the next. run_count 3 averages that out, for three times the credits.
Can I audit for Malaysia or Vietnam?
Yes. market takes indonesia, malaysia, vietnam or global, and language takes id, ms, vi or en, so the questions are written in the language your buyers use.
Connect your assistant