BotRank MCP: 12 GEO use cases for Claude and ChatGPT

Twelve ways to use the BotRank MCP to turn your AI visibility data into ready-to-use reports, briefs and alerts. Each case comes with a prompt to copy and paste, adapted by replacing the bracketed items, and works just as well in Claude as in ChatGPT.

• Ranked by usefulness: the first six are the ones to test first.
• A ready-to-copy prompt for every case, to adapt to your brand.
• Some cases combine with Search Console, GA4 or Slack through n8n to go further.
• The MCP is read-only: no automation can change your prompts or competitors.

The 12 use cases, ranked by usefulness

Each case states who it is for, the prompt to paste into Claude or ChatGPT as is, and the MCP tools involved. Replace the bracketed items with your own values.

1. Automate the weekly AI visibility report [CMO, Agency]

Monday morning, someone opens the dashboard, takes screenshots, pastes them into an email and adds commentary. Two hours, every week. This prompt does the same job in one conversation, in your format.

The prompt to copy:

List my BotRank brands and pick [BRAND]. Compare AI visibility over the last 7 days with the previous 7 days, all engines combined and then engine by engine. Give the 5 prompts that improved most and the 5 that dropped most, with the engine concerned. Flag competitor moves above 2 points of share of voice and any cited domain appearing for the first time. Write a one-page brief, direct tone, for a marketing leadership meeting, ending with three recommended actions.

You get a structured summary you can review in three minutes. For the monthly version, swap "7 days" for "30 days" and ask for a Google Doc or a deck. To stop thinking about it altogether, run the same prompt in n8n every Monday at 8am with automatic delivery.

MCP tools: get-brand-visibility, list-prompts, get-entities-visibility-history, list-sources

2. Write content that gets cited [SEO, Content]

Before writing on a topic, you need to know which pages AI engines already cite on it, what those pages have in common, and what yours is missing. That is half a day of manual research. The MCP brings it down to ten minutes, and the brief it produces rests on real citations rather than a hunch.

The prompt to copy:

On [BRAND], find the prompts containing "[TOPIC]". For those prompts, list the 15 most cited pages over 30 days with their domain. Read the top 5 and extract what they share: structure, angle, length, questions covered, data or tables included. Then fetch the optimization recommendations for my page [MY PAGE URL]. Write a full content brief: title, the angle the cited pages are missing, H2/H3 outline, questions to cover, sources to mention, target length, and the technical fixes to apply.

This works for a new article as well as a rewrite, and it flows naturally into recommendations and the roadmap to track execution. It draws on the same data as source analysis, but turns it straight into a work plan. To get a first draft, add: "Then write the first 600 words from this brief."

MCP tools: list-prompts, list-cited-pages, get-recommendations

3. Spot the pages losing citations, before it shows in traffic [SEO]

A page that ChatGPT stops citing makes no noise. You will see it in GA4 two months from now, once referral traffic has dropped. This prompt tells you this week.

The prompt to copy:

On [BRAND], list every page from my domain [MYSITE.COM] cited by AI engines from [PERIOD 1 START] to [PERIOD 1 END], then from [PERIOD 2 START] to [PERIOD 2 END]. Build a table: URL, citations in period 1, citations in period 2, change. Sort by largest loss. For the three pages losing the most, check which competitor pages replaced them on the same prompts. End with an ordered list of pages to refresh.

The table fits on one screen and the decision is made while reading it. With two-week windows you catch the drop early without reacting to daily noise.

MCP tools: list-cited-pages (two date windows), get-prompt-responses

4. Compare your page with the competitor page the AI prefers [SEO, Content]

On a given prompt, a competitor is cited and you are not. The useful question is not "why them?" but "what does their page have that mine does not?". It is the use case content teams ask us for most, and it fits in one prompt, on any plan.

The prompt to copy:

On [BRAND], take the prompt "[EXACT PROMPT]". Fetch the latest responses from each engine and identify the most cited competitor page. Read that page and mine: [MY PAGE URL]. Compare them point by point: direct answer at the top of the page, heading structure, questions covered, figures and data, freshness, third-party brand mentions, author signals. Add BotRank's optimization recommendations for my page. Conclude with a 5-point rewrite plan, from highest to lowest impact.

One reminder this case makes obvious: a page can be better built and less cited, because off-site authority also weighs in (we explain this in why AI recommends your competitor). If the gap comes from there, the next case is the one to open.

MCP tools: list-prompts, get-last-prompt-responses, list-cited-pages, get-recommendations

5. Get your PR and link-building list sorted by what AI engines actually read [CMO, PR, SEO]

Few marketers know which media outlets AI engines actually read in their market. That list already exists in your data. It only needs sorting.

The prompt to copy:

On [BRAND], list the 100 domains most cited by AI engines over 30 days, with their site type. Produce three separate lists: media and blogs (PR and guest content targets), communities and forums (presence and answers), comparison sites and directories (listings). For each domain, give citation count, share, and whether its share rose or fell over the last 30 days. Mark the ones where [BRAND] is never cited while a competitor is.

You leave with three lists that three different people can work on separately. To prioritize by authority, ask Claude to cross-check with your usual SEO connector (domain authority) or simply paste the authority scores you already have.

MCP tools: list-sources, get-site-citation-share-history, list-cited-pages

6. Produce the competitor share-of-voice table, engine by engine [CMO]

Leadership wants a table, not an animated chart. Who wins on ChatGPT, who wins on AI Overviews, where you are absent, and which new players are rising. This prompt builds it, and reveals your blind spots per engine along the way.

The prompt to copy:

On [BRAND], list brand-type entities over 30 days, for each engine separately: ChatGPT, Gemini, AI Overviews, Perplexity, Claude. Build a table with one row per actor and one column per engine, value = share of voice. Add an "average position" column. Highlight actors not declared as competitors that exceed 3% on at least one engine. End with the three engines where the gap between [BRAND] and the leader is widest, and for each, the prompts where we are absent.

The table pastes straight into a deck. The last part of the prompt, on missing prompts per engine, is often the most useful: that is where the quick wins hide.

MCP tools: list-entities (engine filter), list-prompts (engine filter)

7. Know what ChatGPT says about you without reading 300 responses [CMO, Brand]

Being cited is not enough. A neutral mention while a competitor is presented as "the obvious choice for SMBs" is not a win. The MCP returns the full text of responses, which allows real discourse analysis rather than counting.

The prompt to copy:

On [BRAND], take the 10 most important prompts (tag "[TAG]" or the 10 with highest visibility). Fetch all their responses from the last 30 days. Analyze only the passages that mention [BRAND]: which arguments recur, which adjectives, which competitors we are compared to and on which criteria, which use cases are attributed to us. Compare this discourse with our official message: "[YOUR CANONICAL SENTENCE]". List the gaps, the factual errors, and the phrasings to correct. Output: a two-page brief, per engine.

This is the natural complement to perception and sentiment analysis. The resulting brief feeds directly into your About page, directory listings and press kit.

MCP tools: list-prompts, get-prompt-responses, list-entities

8. Turn Search Console into new prompts to track [SEO]

Your GEO tracking is only as good as your prompt set. Search Console knows what people actually search for. Crossing the two shows what is missing, and catches the flattering prompts that inflate your score.

The prompt to copy:

Here are my top 200 non-brand Search Console queries from the last 3 months, with impressions and clicks: [PASTE EXPORT]. List every tracked prompt on [BRAND]. For each query above 500 impressions with no equivalent prompt, propose a conversational phrasing to add in Prompts Studio. Then audit the existing prompt set: duplicates, prompts that contain the brand name, buying-journey stages not covered (discovery, comparison, decision). Output: a list of prompts to add and a list to remove, with the reason for each.

If Search Console is connected to Claude, replace the paste with "fetch my Search Console queries". The prompts you keep then go into Prompts Studio.

MCP tools: list-prompts, plus your Search Console export or connector

9. Cross AI visibility and GA4 traffic in the same table [CMO]

Most brands track AI visibility without ever tying it to revenue. This prompt is not perfect attribution. It is the first honest answer to the board's question.

The prompt to copy:

On [BRAND], fetch day-by-day AI visibility over 90 days, overall and per engine. Here in parallel are my daily GA4 sessions for the sources chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com, with conversions: [PASTE EXPORT]. Align both series on the same dates. Compute the correlation per engine with a lag of 0, 7 and 14 days. Identify the weeks where visibility moved by more than 5 points and describe what happened to traffic in the following 14 days. Conclude honestly on what the data supports and what it does not.

The word "honestly" is not decorative: ask Claude to say when the relationship is not proven. With an active GA4 connector, no pasting is needed. For what a documented link between AI visibility and revenue looks like, see the Conectic+ case study.

MCP tools: get-brand-visibility (engine filter), list-prompts, plus GA4

10. Quantify a campaign's before and after in five minutes [CMO, PR]

A press release, a launch, a site redesign. Two weeks later, everyone assumes it helped. The MCP lets you measure it on custom dates, including what engines say about you.

The prompt to copy:

On [BRAND], compare two periods: [DATE] to [DATE] (before) and [DATE] to [DATE] (after [EVENT]). For each: overall and per-engine visibility, share of voice against competitors, top 20 cited domains, and the prompts where our average position changed by more than 2 ranks. List the domains newly cited after the event. On the prompts that moved most, compare the response wording before and after. Write a one-page summary: what changed, what is probably linked to the event, what is not.

For PR in particular, the "newly cited domains" line is often the most convincing proof: if the outlet that covered you appears among the sources after the campaign, the effect is visible without a detour.

MCP tools: get-brand-visibility, list-entities, list-sources, list-prompts, get-prompt-responses (custom dates)

11. Produce ten white-label client reports with one prompt [Agency]

The MCP lists every brand your account can access. An agency can therefore loop over its clients and produce one report per brand, in its own branding, in one conversation.

The prompt to copy:

List all my BotRank brands. For each one, produce a monthly report with: AI visibility over 30 days and change, top 3 engines, share of voice against competitors, 5 rising prompts, 5 falling prompts, 10 most cited sources, and three recommendations drawn from the optimization recommendations. Format: one document per brand, titled "AI Visibility Report, [MONTH]", signed [AGENCY NAME], with no mention of third-party tools. Tone: factual, action-oriented. Generate the documents one at a time.

Variant for prospecting: create a trial brand for a prospect, let it run for seven days, and launch the same prompt. You walk into the meeting with a quantified diagnosis, which beats a generic deck.

MCP tools: list-brands, then the tools from case 1 for each brand, get-recommendations

12. Get a Slack alert when a citation drops [Agency, SEO]

The last case does not run in Claude but in n8n. It runs every morning and only bothers you when there is something to see.

The prompt to embed in the n8n workflow:

On [BRAND], for every prompt with the tag "[PRIORITY TAG]", compare visibility over the last 7 days with the previous 7 days. Also compare the citation count of every page on [MYSITE.COM] over the same two windows. If a prompt loses more than 10 points or a page loses more than 30% of its citations, write a Slack message of 5 lines maximum: what dropped, by how much, on which engine, which competitor page seems to have taken its place, and one suggested action. If nothing crosses these thresholds, reply only "NOTHING TO REPORT".

The workflow: a scheduled trigger, an MCP Client node pointing at BotRank, a model node with this prompt, a filter on "NOTHING TO REPORT", a Slack node. Thirty minutes to set up, then nothing left to do. It is the first workflow our agency clients build.

MCP tools: list-prompts, get-prompt-visibility-history, list-cited-pages, through n8n

Six rules for reliable results

1. Always start with "list my brands". The brand ID drives every other call, and Claude finds it on its own once you name the brand.
2. State the time window and the engine. "My visibility" with no detail returns a broad period across all engines, which smooths out the interesting gaps.
3. Do not read a three-day swing as a trend. AI answers move every day; compare whole weeks.
4. Ask for the table before the analysis. A table you can check beats a summary you have to take on faith.
5. Cross with your other connectors (Search Console, GA4, Slack, Google Drive). That is where the MCP goes beyond what the dashboard can do.
6. Keep your best prompts in a shared document. The prompt that works for your brand is a team asset, not the property of whoever wrote it.

FAQ

Which tools does it work with?

Any client that supports the MCP protocol: Claude (web, desktop, Claude Code), ChatGPT, Cursor, n8n, and most compatible agents. The prompts in this article work as is in Claude and in ChatGPT.

Can the MCP change my prompts or competitors?

No. The MCP is read-only. Any change to your configuration happens in the BotRank interface. This is a design choice, so that no automation can alter your tracking without you seeing it.

Does my data leave BotRank?

The MCP sends your AI client only the data your prompts request, on your account, after authentication. BotRank transmits nothing to third parties. What your AI client does next depends on its own terms, which you should check with your provider.

Which use case should I start with?

The weekly report (case 1) if you manage, the content brief (case 2) if you produce. Both give a useful result on the first run, and they teach you how to phrase prompts for all the others.