Insites' own AI Visibility Report looked at audit data across 10,000 local businesses and asked a simple question: what actually determines whether ChatGPT or Perplexity recommends one business over another? The answer only showed up once someone could look at thousands of audits at once rather than one at a time. AI-visible businesses averaged 133 Google reviews against 11 for businesses AI didn't recommend, a 12x gap, and while AI can recognise 93% of businesses by name, only 55% of those come with accurate details attached. The rest, it fills in with confident guesses.
That's not a finding you get from reading one audit. It's a finding you get from running audits at mass scale and asking what the pattern is across all of them. The Insites MCP server is what puts that same kind of analysis inside a Claude conversation, not just for a one-off industry study, but for your own client base, prospect list, or territory.
This post covers what the Insites MCP is, how it connects to Claude, why that's a different way of working than a normal single-audit tool, and eight prompts you can run today.
What is the Insites MCP?
MCP stands for Model Context Protocol, an open standard that lets an AI assistant connect directly to a live data source instead of working from what it already knows. The Insites MCP server is Insites' implementation of that standard: it gives Claude (or ChatGPT, or an automation tool like n8n) direct, real-time access to Insites' audit engine, 70+ website and local SEO checks, and 200+ verified data points per business.
That data quality matters more than it sounds. If AI search engines are only getting accurate details right on 55% of the businesses they recognise, as the Visibility Report found, then an agent working from verified audit data rather than a general web search is starting from a meaningfully better position.
In practice, the MCP connection means Claude can run an audit, pull historic results, benchmark competitors, generate a proposal, or draft outreach, all without you opening the Insites platform. You ask in plain English, Claude calls the MCP server, and the real audit data comes back into the conversation.
It sits alongside Insites' REST API and webhooks as one of three ways to get audit data out of the platform. The API is for developers building custom software, webhooks push data into your stack automatically, and MCP is built specifically for conversational AI agents, no code required.
How it works with Claude
Once MCP is switched on in your Insites account settings, Claude becomes what's called an "MCP client," and Insites is the "MCP server." The connection is a standard OAuth flow, the same kind of permission screen you'd see connecting any app: you authorise Claude to talk to your Insites account, and that's it.
From there:
- You type a request to Claude in plain language, no audit dashboard, no SEO jargon required.
- Claude calls the Insites MCP server behind the scenes to run the audit(s), pull results, or fetch historic data.
- Claude reads the returned data and reasons over it in the conversation, summarising, comparing, ranking, drafting.
- Anything Claude builds, like a matched proposal, links back into your Insites account automatically.
The distinction worth noting: Claude isn't describing SEO best practice from memory. It's interrogating real, current audit data for a specific business, or fifty of them, and telling you what it found.
Why this is different to a traditional audit
A traditional audit tool, Insites included when used the normal way, is built around one workflow: pick a business, run the audit, read the report, repeat. That works fine when you're checking one prospect before a call. It breaks down the moment you want to understand a market, a client base, or a trend, because reading fifty reports by hand and holding the pattern in your head isn't realistic.
Connecting Insites to Claude through MCP removes that one-at-a-time ceiling in four specific ways.
You can audit multiple businesses at once
Ask Claude to "audit 20 dentists in Leeds" and it runs all 20 in one prompt, rather than 20 separate manual audit runs. The output isn't 20 disconnected reports either, it's one ranked view of where the opportunity actually is.
You can audit across multiple audit types
SEO, local SEO, AI visibility, ads, compliance, content and accessibility checks all sit under the same MCP connection. That matters because the Visibility Report found technical SEO factors like Core Web Vitals, page speed and structured data have little correlation with whether AI actually recommends a business, so a single narrow score can miss the signal that's actually driving results.
You can pull several audits over time
Ask for a client's last six months of audit history and Claude can show the trend, not just today's snapshot. The Visibility Report also found that sites which hadn't been updated recently were significantly less likely to appear in AI recommendations, the kind of pattern that's only visible when you're looking at audits across time rather than a single point-in-time check.
You can find the common threads across all of it
This is the one a dashboard genuinely can't do. Claude can look at fifty audits at once and surface the single issue holding back a whole vertical, because it's reasoning across the mass of data rather than presenting one report at a time. The 12x gap in review volume between AI-visible and AI-invisible businesses is exactly this kind of thread: invisible without looking at thousands of audits side by side, obvious once you do.
8 prompts to try
Copy any of these into Claude once your Insites MCP connection is live. Swap in your own city, industry, or client names.
1. Bulk prospecting with a ranked hit-list
"Audit every dentist in Manchester and rank them by biggest opportunity. Flag the ones with the weakest online presence first."
2. Cross-client pattern spotting
"Pull the audits for our 30 most recent clients in the hosting industry. What's the single most common issue across all of them, and what would fixing it be worth in the pitch?"
3. Trend tracking on one account
"Look at every audit we've run on [Client X] over the last six months. Summarise what's improved, what's stayed broken, and write me a two-paragraph update I can send them."
4. Competitive benchmarking at scale
"Audit our top 15 clients in the legal services vertical, then audit their closest local competitor for each. Where's the average gap widest, and where's our retention risk highest?"
5. Territory-level market research
"Audit every accounting firm in Leeds and Bristol. Which city has more low-hanging opportunity, and what's the single issue that shows up most often?"
6. AI visibility gap-finding
"Check how our top 20 clients show up in ChatGPT and Gemini right now. Which ones aren't being recommended at all, and how does their review count compare to the ones that are?"
That last question is worth asking directly, given the 12x review gap between AI-visible and AI-invisible businesses. It's often the fastest way to explain to a client why they're missing from AI search.
7. Post-delivery proof of work
"Re-audit every client we delivered a new website for this quarter. Show me what actually changed compared to the pre-build audit, business by business."
8. Outreach drafted from real findings
"Audit these 10 leads [attach list], match each one to the right service package based on what's actually wrong with their site, and draft a personalised outreach email for each."
Each of these is a single prompt. Getting the same answer manually, running audits one by one and holding the pattern across all of them in your head, is realistically an afternoon of work, if anyone gets around to it at all. For more examples of what an agent can build on top of this connection, see agentic workflows.
Getting started
MCP is available from your Insites account settings, and the connection itself takes a couple of minutes: enable it, authorise Claude, and start prompting. If you're not sure whether it's included on your current plan, it's worth checking with your Insites contact before building a workflow around it.
The tools you'd use one audit at a time still work exactly the same. What changes is that you're no longer the one reading fifty reports to find the insight that mattered.
Coral is Marketing Director at Insites, with more than 10 years leading marketing for B2B and B2C brands. She specialises in brand strategy and growth, building marketing functions that stay aligned with commercial goals and turn marketing investment into measurable results.
