01
Answer coverage
The share of tracked buying questions where an AI engine names the brand.
One number, one direction. Say it as “we appear in 5 of the 30 questions buyers ask” and nobody needs a glossary.
Sophyx MCP connector · for agencies and fractional CMOs
One visibility run. One MCP connector. One report your client reads without you in the room.
The data is already in your workspace. The work is turning it into five pages a CMO acts on, with your logo on it and none of ours. Switch on the Sophyx MCP connector and your own AI assistant does that part — about ten minutes per client, per month.
Free to start · no credit card · run it on a prospect before they sign
One run in, one branded client report out
Walkthrough video landing shortly
What metrics should I use to report AEO results to leadership?
Five, in this order. Answer coverage — the share of tracked buying questions where AI names the brand. Who is named instead — the competing brands the engines recommend in your place. Gaps closed this period — the unit of work you invoiced for. Citation sources — the pages and domains the engine built its answer from. And the coverage trend over six weeks minimum, because a single month is noise. Do not headline position-in-answer: assistants are probabilistic, the rank moves between runs, and your client’s second tool will report a different one.
AI visibility tools produce numbers that are true and unusable
A coverage percentage means nothing to a client who has never heard of coverage. A list of prompts is a list. And every tool in this category reports a different number, which your client will notice the moment they get a second opinion — usually in front of someone you were trying to impress.
A report that survives a leadership meeting does four things: it says whether the brand shows up, it says who is showing up instead, it says what closing the gap takes, and it says what changed since last month.
Everything else is decoration:
- A coverage percentage with no sentence attached to it.
- Position-in-answer as a headline, against a probabilistic system that moves between runs.
- A list of prompts with no verdict on any of them.
- Anything the client would have to look up before they could react to it.
The five that survive the meeting
In priority order. If the room only has time for one, it is the first.
02
Who is named instead
The competing brands the engines recommend in your client’s place, ranked by frequency.
The competitive fact. It creates urgency without you having to manufacture any.
03
Gaps closed this period
How many previously-absent questions now return the brand.
The unit of work. It maps directly to what you invoiced for, which is the number a retainer is renewed on.
04
Citation sources
The pages and domains the engine actually built its answer from.
It tells you what to build next, and it turns “do more content” into a specific list.
05
Coverage trend, six weeks minimum
The same prompt set, re-run on a fixed schedule.
A single month is noise. Six weeks is a direction, and a direction is what a board actually buys.
What “white label” actually has to mean
Agencies tell us a logo on a PDF is not white label. It has to be the whole artefact: your name, your language, your structure, your recommendations — and no product terminology your client can search for and find priced publicly.
Your Agency
AI visibility report · August
16.7%
answer coverage
The brand appears in 5 of the 30 questions its buyers ask. One sentence, no jargon, and the client knows where they stand before they reach the second page.
recommended instead
SemrushProfoundPeec AIOtterly.AIAhrefs
- 01Answer coverage — 5 of 30 buyer questions name the brand
- 02Who AI recommends instead, ranked by how often
- 03Three gaps to close this month, with the hours each takes
- 04The trend line, six weeks minimum, so it reads as a direction
prepared by Your Agencypowered by Sophyx
Because the assistant composes the report from the raw data rather than exporting a template, it is your document from the first draft. The figures above are our own — 30 tracked prompts, 5 cited, 16.7% coverage, and the brands the engines named in our place.
Add the MCP connector to the assistant you already pay for
About four minutes, once, for every client workspace you will ever run.
Server address
https://mcp.sophyx.app/
The same address works in ChatGPT, Claude, Microsoft Copilot and Gemini. The steps below are ChatGPT’s, because it is the only one that hides the feature behind a toggle.
- 01
Open Settings → Connectors
Your name, bottom-left in the desktop app, or your avatar top-right on the web. Depending on your build the panel reads Connectors, Apps & Connectors, or Plugins — it is the same panel.
- 02
Scroll to the bottom and switch on Developer mode
It is the very last item. Do this before anything else.
Until this is on there is no add button anywhere, and searching Settings for “MCP” returns nothing. It is not that you are missing it — it genuinely is not rendered yet.
- 03
Add the connector
Browse connectors → scroll to the top → the + button to the right of the search bar. Name it Sophyx, paste the server address, choose OAuth, tick the trust confirmation, Create.
- 04
Approve with your Sophyx login
Sign in, scroll to the bottom of the permission screen, Allow access. That window often opens behind the app — check your taskbar if nothing seems to happen. Then tick Sophyx in the composer’s + menu and you are live.
Developer mode is a paid-plan feature. On ChatGPT Free there is no Connectors panel at all. Claude, Copilot and Gemini need no equivalent toggle — paste the address and approve.
Two prompts: the retainer report, and the pitch
Take them. Same data, two rooms, and the second one is where the money is.
Prompt 1 — the monthly client report
paste into ChatGPT or Claude
Using the Sophyx connector, build a client-ready AI visibility report for
[workspace].
Structure:
1. One sentence a non-marketer understands: is this brand showing up in AI answers or not
2. Answer coverage: the number, what it counts, and what it means in plain English
3. Who AI recommends instead — the competitor brands named across the gaps, by frequency
4. The three highest-value gaps, with what closing each one would take in hours and assets
5. A 30-day plan with named deliverables and owners
6. What we will re-measure, and when
Rules: no jargon, no metric a client would have to look up, no claim the data does not
support. Where something is uncertain, say so in one line rather than dropping it.
Brand it as [agency name]. Output as a standalone HTML file I can print to PDF.Replace the two bracketed fields, review once, print to PDF. About ten minutes.
Prompt 2 — the version you read before a pitch
paste into ChatGPT or Claude
Using the Sophyx connector, run the pitch version for [prospect brand].
Give me:
- The one uncomfortable sentence: which buying questions AI answers without them
- The competitors AI recommends in their place, with how often
- The single fastest gap to close, and roughly what it would take
- Three questions I should ask them in the meeting that their own data now makes obvious
Keep it to one page. I am reading this in a meeting, not sending it.Run it on a prospect's brand. The free audit covers any domain, so this costs nothing to test.
The prompts are free and always will be. They only compose a real report once there is a tracked workspace behind them.
Create your first client workspaceThe pitch use is worth more than the retainer use
Run the analysis on a prospect’s brand before the meeting. You arrive knowing which competitors AI recommends in their category and which questions they lose — a specific, checkable, uncomfortable fact about their business that nobody else in the room has.
It reframes the meeting from “should we do this” to “how bad is it”, and you did not have to sell a category to get there.
What you get per client, every month
All of it included. None of it a separate line item.
Every client brand in its own workspace
Separate prompts, separate competitors, separate trend. Run as many as you have clients — the connector reads every workspace on your account in one conversation.
The report written from raw data, not exported from a template
Because your assistant composes it, it is your document from the first draft — your structure, your language, your recommendations.
Your mark, and none of ours
No product terminology your client can search for and find priced publicly. A logo on a PDF is not white label, and agencies tell us so.
A different report for a different room
A retainer client gets the trend. A pitch gets the competitive gap. Same data, same ten minutes, two documents.
The prospect version, before they sign
Run the analysis on a brand you are pitching and arrive holding a specific, checkable fact about their category that nobody else in the room has.
From USD 39 per month, and the first audit on any domain is free. One report that wins one retainer pays for the year. See pricing.
White-label AI visibility reporting — your questions answered
- Five, in this order. Answer coverage — the share of tracked buying questions where AI names the brand. Who is named instead — the competing brands the engines recommend in your place. Gaps closed this period — the unit of work you invoiced for. Citation sources — the pages and domains the engine built its answer from. And the coverage trend over six weeks minimum, because a single month is noise. Do not headline position-in-answer: assistants are probabilistic, the rank moves between runs, and a second tool will report a different one.
- The Sophyx MCP connector is a Model Context Protocol server at mcp.sophyx.app that lets ChatGPT, Claude, Microsoft Copilot or Gemini read your Sophyx workspaces directly. It matters for reporting because it is the difference between exporting a template and composing a document: your assistant reads the raw run — coverage, gaps, competing brands, citation sources — and writes the report in your structure and your language. That is what makes it yours rather than ours with a logo swapped in. It reads every workspace on your account, so a multi-client agency sets it up once.
- Yes. Connected over MCP, your AI assistant writes the report from your workspace data as a document you own, under your agency's name, in your own structure and language. Because it is composed rather than exported, there is no product terminology left in it for a client to search and find priced publicly.
- About ten minutes: one prompt, one review pass, print to PDF. The analysis itself runs inside Sophyx on whatever schedule you set, so the data is already waiting when you sit down to write.
- Monthly for the report, and six weeks minimum before you read a trend into it. Anything shorter is noise dressed as movement, and reporting noise to a client every fortnight trains them to distrust the whole measurement.
- That every tool tracks a different prompt set against a probabilistic system, so the absolute number is not comparable between tools — only your own number over time against a fixed prompt set is. Report the delta and name the prompts. Whoever makes their number defensible wins that conversation, and defensibility comes from a stable prompt set, not from a bigger figure.
- Each brand gets its own workspace with its own prompts, competitors and trend, and Sophyx is built for agencies running several at once. Plans start at USD 39 per month and the first audit on any domain is free, which is also the cheapest way to test the pitch version on a prospect before you commit anything.
Put a report in front of a client this month
Create a workspace for one brand, connect your assistant, and the first report is about ten minutes of work. Start with the prospect you most want to win.
Explore next
Turn the report into a build listCross the client’s invisible prompts against the searches they already get impressions for.Ship the pages that close the gapsConnect your assistant and publish straight to the client’s site.Start freeCreate a client workspace and build the first report in about ten minutes.