Investigate AI Visibility with SEO MCP

Use this workflow when a project-level AI Visibility chart changes or a brand appears in an AI answer. The chart identifies where to look; the stored prompt and answer show what was actually observed. Cite a specific provider, market, and analysis date because generated answers can vary.

Confirm the dataset

Call list_projects, then get_ai_visibility_options for the selected project. Check configured providers and locales, last analysis time, and prompt usage. If the requested provider or market was not tracked, report that limit rather than interpreting an empty list as a negative result.

Move from overview to evidence

1. Use get_ai_visibility_overview to identify the observed trend, recent mentions, or cited-source pattern.
2. Use list_ai_visibility_prompts or list_ai_visibility_mentions to find relevant tracked records. Follow the returned cursor when the first page does not cover the question.
3. Open one returned answer with get_ai_visibility_answer. Read the actual mention, sentiment statement, and citations together.
4. Use list_ai_visibility_sources and source or prompt activity tools when you need to trace recurring domains or pages.
5. Report what was observed, what is missing, and which content or technical question should be checked next. A citation is not proof that an edit caused the answer.

For [PROJECT], investigate the stored AI Visibility change for [BRAND] in
[MARKET]. Name the provider and dates. Open the underlying prompt and answer,
then list the cited source pages and the exact observed mention. Distinguish
the answer from an inference. Do not add or delete prompts.

Decide the next useful check

If a cited page belongs to your site, inspect its current public content and technical accessibility before proposing an edit. If a competitor is cited, compare the answer's specific claim with reliable sources; do not copy its wording. The content and AI search audit guide covers technical foundations without promising future citations.

Adding or deleting prompts is a separate, capacity-sensitive decision. The MCP server cannot export AI Visibility files, suggest prompts, or manually reanalyze an existing prompt. Read AI Visibility tools for the exact data and action boundaries.