The OtterlyAI MCP connects your AI search data and agent analytics to Claude, ChatGPT and any other MCP-capable assistant. Once it’s connected, you can chat with your data, build reports and dashboards, and run multi-step GEO workflows in plain language.
The question we hear most often from customers is a simple one: What are the best prompts to use with the OtterlyAI MCP? So this post collects 12 prompts you can copy right now. The first six are the ones I share most often with customers. The remaining five are the most-used workflows in the OtterlyAI Marketplace, where you’ll find 100+ prompts in total.
Every prompt uses [BRACKETS] for the values you need to replace: your brand report ID, your domain, your country.
Before you start: connect the MCP and find your brand report ID
You need two things: the OtterlyAI MCP connected to Claude (or ChatGPT) and the ID of the brand report you want to query.
Connect the OtterlyAI MCP in Claude. Open Settings, go to Connectors, and add OtterlyAI. You’ll need an OtterlyAI API key, which you can generate in your account settings. The Help Center has a step-by-step walkthrough.
Find your brand report ID. Open the brand report in OtterlyAI; the ID is the last part of the URL. If you’re not sure, you can also just ask Claude: “List my OtterlyAI brand reports” and it will return the names and IDs.
Paste a prompt below, replace the bracketed values, and run it.
Tip: Claude keeps context within a conversation, so once you’ve named the brand report once, follow-up prompts in the same chat don’t need to repeat it.
1. See your underperforming prompts
The fastest way to get a to-do list out of OtterlyAI. Claude pulls the prompts with the lowest brand coverage and pairs each one with on-page and off-page recommendations.
Use the OtterlyAI MCP
Brand report ID: [BRAND REPORT ID]
Give me a list of the underperforming prompts and recommend what I should do on-page and off-page.
2. Build a custom domain dashboard
This is my personal favorite. It combines AI search data with agent analytics (if you’ve installed OtterlyAI on your website) and shows the gaps between agent crawls, visits and AI citations. On otterly.ai, for example, AI agents made 33,836 visits to 56 cited URLs over a 30-day window, while 26 cited URLs got no agent visit at all.
Use the OtterlyAI MCP
Brand report ID: [BRAND REPORT ID]
Can you create a domain dashboard for [my domain], putting citation data next to agent analytics data? Calculate a conversion rate (agent visits per citation) for each URL of [my domain].
3. Find prompt gaps with Google Search Console
Match the question-style queries you already rank for on Google against the prompts you’re tracking in OtterlyAI. The output is a table of GEO opportunities where you rank well on Google but are missing in AI answers. Export your GSC queries as a CSV and attach it, or use a connected GSC source.
Using my Google Search Console data (attached file) and my OtterlyAI brand report [BRAND REPORT ID], do the following:
1. Pull the top 20–30 question-style queries I already rank for in GSC (last 90 days). Filter for natural-language patterns: how, what, why, best, vs, alternative, top, should I, is X worth, X for Y. Prioritize by impressions × position (rank ≤ 20).
2. Classify intent for each query: Informational / Commercial / Navigational / Transactional.
3. Convert each query into a natural-language AI prompt, the way a real user would phrase it to ChatGPT, Perplexity, Gemini or Google AI Mode. Keep the intent intact but make it conversational (e.g. "best CRM for startups" → "What's the best CRM for an early-stage startup in 2026?").
4. Cross-reference with OtterlyAI: for each generated prompt, check whether OtterlyAI is already tracking an equivalent prompt for my brand. Mark it Tracked (with a link to the prompt) or New (recommend adding). For tracked prompts, pull the current brand-mention rate and NSS.
5. Output a single table with columns: GSC Query | Impressions | Avg Position | Intent | AI Prompt | OtterlyAI Status | Brand Mention % | NSS | Priority.
End with an "Overlap Gap" section: the prompts where I rank well on Google (position ≤ 10) but have zero or low brand presence in AI engines. Rank them by GSC impressions and recommend the top 5 to add to OtterlyAI tracking this week.
4. Create content briefs from citation gaps
Find the prompts where your site isn’t cited, see who is being cited instead, and get a brief per prompt built from the patterns of the winning sources.
Find my weakest prompts in brand report [BRAND REPORT ID] (lowest brand coverage AND lowest domain coverage) for [my brand / my domain] in [country] over the last 30 days.
For the bottom 5, pull the citation report so I can see which domains AI is citing instead of mine. Then write a complete content brief for each prompt:
- Recommended URL slug
- Meta description (≤155 chars, written to be quoted by AI)
- H1 and suggested H2 structure, mirrored from the patterns of the top-cited sources
- The 3 most common domain categories appearing in citations for that prompt
- 3 recommended internal links from my own domain if it appears in the citation data, otherwise logical internal link anchors based on the prompt topic
- One-sentence "AI answer hook": the opening sentence my content should contain so AI is likely to quote it as a direct answer
Format as one brief per prompt, clearly separated.
5. Audit your Wikipedia opportunities
Not sure whether you should invest in a Wikipedia page? Run this first. It checks which Wikipedia URLs AI engines cite in your category and whether your brand appears on any of them.
Based on brand report ID: [BRAND REPORT ID]
Analyze how visible [my brand] is on Wikipedia.
Analyze the most cited Wikipedia URLs in the last 30 days in [country].
Check if my brand is mentioned on those URLs.
Check if there are any recommendations provided by OtterlyAI.
Put together an executive summary on how we should optimize our Wikipedia presence.
Turn this into a Wikipedia-style report with charts in the style of wikipedia.org.
6. Use query fan-outs to find content gaps
AI engines break a prompt into several sub-queries before answering. This prompt extracts those fan-outs for your top prompts and checks them against your sitemap, so you see which sub-questions you have no page for.
Use the OtterlyAI MCP
Brand report ID: [BRAND REPORT ID]
Extract the query fan-outs for our top prompts, then check them against our website sitemap ([my domain]). Which queries do the AI search engines use that we don't have a URL or page for?
7. Compare share of voice against your top 3 competitors
The most-requested competitive prompt in the marketplace. You get a ranked table with share of voice, brand coverage and NSS per competitor, and a delta against the prior period if you ask for it. Run it weekly. Marketplace page
Compare share of voice of my brand to [brand1], [brand2], and [brand3] in OtterlyAI brand report [BRAND REPORT ID] for [month, year], [country]. Show the delta vs the prior month.
8. Find prompts where you’re invisible but a competitor dominates
This is the shortlist for your next content sprint: every prompt where your brand isn’t mentioned but competitors are, sorted by volume. Marketplace page
Based on my OtterlyAI brand report [BRAND REPORT ID], list all the prompts where my brand isn't mentioned but my competitors are, sorted by volume. [Country], last 30 days. For each prompt, name the competitor that is mentioned most often.
9. Check your citation gap: mentioned, but not linked
Being talked about is not the same as being cited. This prompt compares brand coverage with domain coverage and lists the prompts where AI names you but never links to your site. Marketplace page
What's our brand coverage vs domain coverage gap in OtterlyAI brand report [BRAND REPORT ID] for the last 30 days, [country]? List the prompts where my brand is mentioned but none of my URLs are cited, and recommend how to close the gap.
10. Read what ChatGPT actually said about you, and flag outdated claims
Stats tell you whether you’re mentioned. The raw responses tell you how. The first prompt pulls the literal text; the second scans for stale or wrong claims about your brand, which is a monthly must for PR and brand teams. Marketplace page · Misinformation check
Pull the latest 3 ChatGPT responses for the prompt "[prompt query]" from OtterlyAI brand report [BRAND REPORT ID]. Quote the passages that mention my brand and list the citations in each response.
Using OtterlyAI brand report [BRAND REPORT ID], scan the last 20 ChatGPT and Perplexity responses mentioning [my brand] and flag any factually outdated or incorrect claims. For each one, give me the quote, the engine, the date and a suggested correction.
11. Run a crawlability check on a key landing page
None of the above matters if AI engines can’t fetch your pages. This prompt triggers an on-demand crawlability check against GPTBot, ClaudeBot, PerplexityBot, Google-Extended, OAI-SearchBot, ChatGPT-User and CCBot, and returns a per-bot table of robots.txt rules and live HTTP status. Run it on every launch. Marketplace page
Run an OtterlyAI crawlability check on [URL list] and tell me which AI bots are blocked, why, and how to fix it.
12. Create new brand reports and prompts directly via MCP
You don’t have to leave Claude to set up tracking. The OtterlyAI MCP can create a brand report, add prompts to it and set the country in one go. Handy for agencies onboarding a new client, or for spinning up a quick test report before you commit prompts to your main one.
Use the OtterlyAI MCP
Create a new brand report with the following details:
Brand name: "[brand name]"
Report name: "[report name]"
Website: [my domain]
Then create and add the following prompts to it:
"[prompt 1]"
"[prompt 2]"
Country: [country]
Claude creates the report, adds the prompt and returns the new brand report ID, which you can use in any of the prompts above.
How to write your own prompts
After watching hundreds of customers use the MCP, the prompts that work share four traits.
Name the brand report, country and time window up front. “Last 30 days, US” removes most of the ambiguity Claude would otherwise have to guess at.
Ask for one output shape. A table with named columns, a ranked list, a dashboard. Claude is much better at filling a structure than inventing one.
Chain the steps. The GSC and content brief prompts above work because each step feeds the next: pull, filter, cross-reference, output. Number the steps and Claude will follow them in order.
End with the decision. “Recommend the top 5 to add this week” turns a report into a to-do list. Reports without a next step get read once and forgotten.
70+ more prompts in the OtterlyAI Marketplace
These 12 are a starting point. The OtterlyAI Marketplace holds 100+ prompts, agents, tools and resources for GEO, organized by job: prompt research, reporting, citations, sentiment, automation, competitive analysis and more. Many run on a schedule and push results into Slack or your CMS.
If you want help setting up AI search monitoring for your own business, book a demo with one of our experts, or start a free trial and paste the first prompt into Claude today.
