Skip to main content
Conductor’s MCP server connects your Conductor account to AI assistants (such as ChatGPT or Claude) through connectors. You ask questions in natural language; the assistant uses MCP tools to read allowed Conductor data and answer from your metrics—not generic SEO advice. The MCP spans three broad areas of your Conductor data: AI search (how your brand shows up in AI-generated answers), traditional search (how your site ranks in classic search engine results), and page performance (how pages perform across traffic, AI citations, bot activity, and technical health). You can analyze each on its own or ask the assistant to connect them—for example, comparing where you rank traditionally against where you’re mentioned and cited in AI search, or checking whether a page that lost traffic also changed. A fourth area—content creation—lets you create and change content in your account—running the Writing Assistant workflow from your assistant. See Content creation below. For connector setup, use the Getting Started pages in this MCP section.

​
What kinds of questions can you ask?

The conversational interface is meant to help you probe your data and surface insights you might not have clicked to in the product. For example:
What are the topics in AI search where my site is cited most? Which pages are cited the most? How am I performing on my tracked prompts compared to competitors? What topics should I prioritize for content optimization? What topics might I consider tracking in Conductor that I do not track yet?
For traditional search, you can probe rankings, demand, and the SERP itself:
Which of my tracked keywords improved or declined in rank over the last quarter, and which keyword groups are driving the change? Where are competitors outranking me, and which keywords show the biggest gaps? Which keywords are seasonal, and when should I publish to catch the next demand peak? For my priority keywords, which SERP features (People Also Ask, Local Pack, and others) appear, and how should that change my content format?
You can iterate from there—the assistant can help you refine questions as you learn what the tools return.

​
Tools and analysis areas

The MCP exposes focused tools (named capabilities the model can call) that align with datasets in Conductor Intelligence, across AI search, traditional search, and page performance:

​
AI search analysis

These capabilities focus on how your brand (and competitors) show up in AI-generated answers across supported AI search engines.

​
Brand mention analysis

Tools and prompts here focus on how your brand (and competitors) show up in AI search performance data—often discussed as brand mentions. Useful angles include:
  • Topics and prompts: Which topics or queries (in the data, a tracked prompt is treated as a query) drive the most brand mentions for you, including unbranded queries where you still appear.
  • Competition: Share of voice vs competitors, queries where you lead or trail, simple charts (for example pie charts) when the assistant can build them from tool output.
  • Over time: How brand-mention share of voice has trended.
  • Persona and intent: Strongest and weakest areas by persona or search intent (as modeled in Conductor).
When you ask, use the phrase “brand mention” so the model routes to the right tools and framing. You can also ask whether actual AI response snippets are available where you have brand mentions; that slice of the experience may still be coming online depending on your program version—confirm with your assistant or Conductor contact if you need raw snippet text.
Example prompt: “Which topics and queries drive the most brand mentions for us, and where do competitors out-mention us? Show share of voice across the top topics.”

​
Citation analysis

A separate set of capabilities focuses on citations: which URLs and domains AI engines cite as sources. You may see tool names such as ai_citations_marketshare in explanations or logs. Typical questions include citation share vs competitors, topics or queries where you are mentioned but not cited, performance over time, and breakdowns by persona or intent.
Example prompt: “Which of our pages and domains earn the most citations in AI answers, and which topics mention us but don’t cite us as a source?”

​
Sentiment analysis

Sentiment tooling helps you understand how AI responses talk about your brand in connection with tracked prompts. Sentiment analysis applies to your own AI responses within the supported scope; you can still ask the assistant to compare or contextualize you vs competitors for brand mentions in other flows. Prompts often combine topics, queries, time trends, persona, and intent.
Example prompt: “How do AI responses talk about our brand across tracked prompts—what’s the positive, neutral, and negative split, and which topics carry the most negative sentiment?”

​
Traditional search analysis

These capabilities focus on how your site performs in classic search engine results for your tracked keywords. Use the phrase “traditional search” (or “keyword rankings”) so the model routes to the right tools.

​
Rankings and visibility

Ask where you rank and whether it’s improving—across your whole keyword set or any slice of it. Useful angles include:
  • Rank trends: How rankings have moved over time for individual keywords or keyword groups, and where you’ve won or lost visibility.
  • Segments: Performance by keyword group or location/market, so you can prioritize the keywords and regions that need attention.
Example prompts:
  • “How has our visibility moved over the last 8 weeks across our priority keyword groups? Highlight any group with a Top-10 drop.”
  • “Using the rank opportunities preset, give me up to 20 keywords currently on page 2 in our ‘product’ keyword group.”
  • “Break our visibility down by tracked location for the last 4 weeks. Which markets are pulling us up and which are dragging us down?”

​
Seasonality and demand

Ask about monthly search volume trends (typically the last 24 months) to separate ranking changes from shifts in underlying demand. This helps you explain why visibility or traffic moved and time content and campaigns to seasonal peaks.
Example prompt: “Which of our tracked keywords are most seasonal, and when does demand peak—so we can time content to the next upswing?”

​
Result types (SERP features)

Ask which SERP features make up a keyword’s results—Standard Links, People Also Ask, Local Pack, and others. This explains why ranking well doesn’t always drive clicks and points to concrete content-format recommendations.
Example prompt: “For our top product category, which SERP features do we own vs. don’t own—and which 3 should we go after first?”

​
Competitive rankings

Ask how you stack up head-to-head against competitor domains (up to three) for your tracked keywords. The assistant can identify where rivals are pulling ahead and ground gap analysis in real position data.
Example prompt: “Run a rank comparison against [Competitor A], [Competitor B], and [Competitor C] for last week. Where are we losing the most ground?”

​
Keyword deep dives

Ask about a single keyword to move from account-wide summaries to specifics: a rank snapshot, week-by-week rank history, monthly search volume trend, and the full SERP for that term.
Example prompt: “Show me the full picture on ‘[exact phrase]‘—rank history, seasonality, and what’s currently ranking on the SERP.”

​
Page performance analysis

BetaThis feature is in a Beta program. Aspects of it are subject to change with little or no notice.
Page performance tools require Conductor Intelligence with Pages data. They’re available in Conductor’s native apps for ChatGPT and Claude, and through custom connections to the v4 connector. The Copilot agent and app don’t include them yet, but you can use them in Copilot through a custom connection.
These capabilities bring the data behind the Pages report into your assistant, so you can connect how a page performs to the likely reason why. Use the words “pages” or “page group” so the model routes to the right tools. Useful angles include:
  • Traffic: Organic and AI referral sessions over time—by page group, website, web property group, or AI engine—and AI vs. organic for the same set of pages.
  • Citations: Which pages AI engines cite most, which engines cite them, and the prompt topics they’re cited for—or pages that earn impressions but no citations.
  • Bot activity: Which bots and bot types visit your pages most, and how that changes over time.
  • Changes: Which pages changed their title, meta description, H1, status, indexability, or canonical, and what happened to their traffic afterward.
Example prompts:
  • “Which pages in our Blog page group lost the most visits last week, and did anything change on them?”
  • “Which pages get Search Console impressions but no AI citations?”
  • “Which AI bots visited our product pages most in the last four weeks?”
  • “Summarize our Product Pages page group: top pages by traffic, most cited, and most visited by bots.”
Scope your questions up front. Name a page group and a date range so your first answer is a list, not a follow-up question. When more than 100 pages match, the assistant tells you how many and asks you to narrow down.

​
Content creation

BetaThis feature is in a Beta program. Aspects of it are subject to change with little or no notice.
Content creation tools are available in Conductor’s native apps for ChatGPT and Claude, and through custom connections to the v4 connector. The Copilot agent and app don’t include them yet, but you can use them in Copilot through a custom connection.
Content tools run the Writing Assistant workflow in conversation: build a brief, generate a draft, refine it, and score the result—against your account, your Content Profiles, and your own research.
Unlike every other tool in Conductor’s MCP, these tools create and change content in your account. They can save drafts, build briefs, and write edits, which means they can consume your organization’s allotted drafts. The MCP tells you when an action uses a new draft so you can watch your usage.
Useful angles include:
Build me a brief for a page targeting “enterprise content governance” on our main web property, using our B2B Content Profile. Score this live page and tell me what’s weakest before we decide whether to rewrite it or leave it. Generate a draft from the brief we just approved, then suggest title tags and meta descriptions. Which of my drafts from the last month haven’t been generated yet? Suggest internal links for this draft and explain why each one fits.

​
How the workflow fits together

Every piece of content—net-new, a live page you want to improve, or unpublished HTML—enters the same way: through a brief. Every meaningful step stops for your approval.
1

Collect the inputs

The assistant gathers the source content, the account and web property, the search locale, up to five topics, and—optionally—a Content Profile, draft title, and any Knowledge Sources. It repeats that summary back to you unchanged.
2

Approve the brief

On your confirmation, the brief is created and researched. If you pointed it at an existing URL or source HTML, that source is scored too, so you can see where it stands. It stops there and waits for your feedback.
3

Review the brief and the score

Read the brief’s insights, adjust them if you want, and look at the baseline score before generating anything.
4

Approve generation

Generation asks for explicit approval first. The assistant then writes the draft, an outline, or a revision of what already exists.
5

Refine and save

Ask for title, meta description, or internal-link suggestions, then save the ones you choose back to the draft.
Long-running steps are asynchronous—the assistant polls for progress and reads results only once the work has succeeded. When a brief finishes, it returns a link that opens that same draft in Writing Assistant, so you can pick the work up in the product at any point.
Ask the assistant to show you the brief and the baseline score before you approve generation. That is the best moment to change direction: the topics, the Content Profile, and the sources are all still adjustable, and nothing has been written yet.

​
What content tools will not do

  • No deletion. Drafts, Content Profiles, and Knowledge Sources cannot be deleted through the MCP. Delete them in the Conductor platform.
  • No file upload. Knowledge Sources can be created from plain text only—not files, paths, or encoded file contents. Sources you uploaded in the platform are still readable here; their files just are not transported through the MCP.
  • No bulk operations. Suggestion tools are scoped to one draft per call. An assistant can loop over several drafts, but there is no batch endpoint—for those use cases, consider the Content API.
  • No configuration writes except for Content Profiles.
  • No automatic internal linking. Link suggestions are review-first: repeated anchor text makes blind insertion unsafe, so you approve the ones you want before they are written in.

​
Account configuration

You can let the LLM review your current configuration—topics, prompts, brands, competitors, personas, intents, locales, search engines, and tracked keywords—to discover gaps or review your tracking strategy across both AI and traditional search.

​
Discovery prompts (learn what you have access to)

Before deep analysis, you can inventory what the connection can see—still in plain language:
  • Show me which Conductor accounts I have access to.
  • Use [account name and ID] for further analysis.
  • Show me the tools and data I have available via Conductor.
  • What time ranges does Conductor data cover for me?
These reduce guesswork before you dig into deeper analysis.

​
Use vocabulary Conductor already uses

Vague words like “visibility” are easy for humans but ambiguous for tools. Prefer product language. For AI search: brand mentions, citations, topics, queries (for tracked prompts), personas, intents, and competitors by name when comparing. For traditional search: keywords, keyword groups, rankings (or rank), search volume, seasonality, SERP features (such as People Also Ask and Local Pack), and locations. For page performance: pages, page groups, web property groups, sessions, AI citations, bot visits, and indexability. For content creation: brief, draft, Content Profile, Knowledge Source, content score, internal links, title tag, and meta description—rather than vague terms like “the doc” or “the SEO settings.” You will get clearer tool use and answers. Name the account and web property when you have more than one. The Conductor glossary defines these and the rest of the platform’s vocabulary.

​
What to keep in mind

  • Access and privacy follow your org’s rules; tools only return what your account is allowed to see.
  • Validate important decisions; the assistant summarizes and explores—it does not replace your judgment or governance.
  • Approve before you generate. Content tools write to your account and can consume allotted drafts. Review the brief and baseline score before approving generation.
  • Review our template gallery for ideas of how to apply the MCP with LLM skills to deepen the analysis and insights.
  • Ask your Conductor team for updated prompt examples, recommendation-style wrap-ups, and citation-specific patterns.