Step 1
Pull the gaps
Every tracked prompt where your brand did not appear, and the competitors named in your place. This is demand written in your buyer’s own words, not in keyword shorthand.
One report lists the prompts AI answers without naming you. The other lists the searches Google already shows you for and nobody clicks. Where they overlap is a page worth building — and once the Sophyx MCP connector is switched on in ChatGPT or Claude, your own assistant finds every overlap and ranks them in about ten minutes.
Free to start · no credit card · your first run finishes in about a minute
Gap list to ranked build plan, in one conversation
Walkthrough video landing shortly
Close them one question at a time, in demand order. List the prompts buyers actually ask, run them against ChatGPT, Gemini, Claude, Perplexity and Copilot, and record every answer your brand is absent from — that is your gap list. Then pull your Search Console queries for the same period and match each gap to the query that means the same thing. A gap with impressions behind it is a page worth building; a gap with nothing behind it is a note for next quarter. Publish a page that answers that exact question in plain, quotable sentences with FAQ and schema markup, then re-run the same prompt four to six weeks later and check whether the answer changed.
Not money — time, which you cannot get back. Here is what the guessing version actually looks like from the inside:
The fix is not more content. It is knowing, before you write anything, which single question is costing you the most — and being able to prove it.
Our own visibility run found 25 gaps against 30 tracked prompts — 16.7% coverage. These are four of them, crossed against our own Search Console data. We publish our losing rows because a method you cannot check is a claim, not a method.
sophyx.io · visibility run 30 Jul 2026 · Search Console, 90 days to 20 Aug 2026
4
gaps matched
3,005
impressions behind them
2
worth building
Best AI tracking tools for brand visibility?
Google: ai brand visibility software
2,325
impr.
40.8
pos.
How can I track brand mentions inside AI answers?
Google: ai brand mentions
566
impr.
70.8
pos.
Monitoring ChatGPT brand mentions tools
Google: free tools to track brand mentions in chatgpt
112
impr.
2.7
pos.
What metrics should I report to leadership?
Google: ai brand visibility report
2
impr.
70.5
pos.
Read row one. AI will not name us when someone asks for the best brand-visibility tracking tools, and Google shows us 2,325 times a quarter for the commercial query that means the same thing — at position 40.8, which earns nothing. Two systems, same verdict, same buyer. That is not a hypothesis. That is a brief.
Run it once a month. It takes about ten minutes once your prompts are tracked.
Step 1
Every tracked prompt where your brand did not appear, and the competitors named in your place. This is demand written in your buyer’s own words, not in keyword shorthand.
Step 2
Your Search Console queries for the same 90 days. Ignore clicks at first — impressions sitting at position 20 and worse are the interesting ones: proven demand, unearned attention.
Step 3
Match each gap to the query that means the same thing, and score the match on impressions, position, and whether the person asking is close to buying.
Step 4
Demand and no page → BUILD. Demand and a page ranking below 20 → OPTIMISE. No demand → WATCH, and revisit next quarter. Now your content calendar is evidence instead of opinion.
About four minutes, and one step that stops almost everybody. No API key, nothing to install.
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.
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.
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.
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.
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.
Take it. It is the entire process, and it works the moment a workspace is connected.
The gap analysis prompt
Using the Sophyx connector, pull my latest AI visibility run and list every
prompt where my brand was NOT found, with the competitor brands named instead.
Then take my Search Console query export (pasted below) and match each gap to the closest
real query, showing impressions, average position and clicks.
Return a table: gap prompt | matched query | impressions | position | verdict.
Verdict is BUILD (demand, no page), OPTIMISE (demand, page ranking below 20) or WATCH (no
demand yet). Sort so BUILD rows come first.
Then write the brief for the top BUILD row: the exact question the page must answer in its
first paragraph, the H2s, and the six FAQ questions.Sophyx supplies the visibility run directly. The Search Console export takes about thirty seconds: Performance → Queries → Export.
The prompt is free and it always will be. It only returns anything once there is a tracked workspace behind it — which is the part Sophyx does.
Create the workspace freeNot a dashboard. A decision, and the work that follows from it.
Real buyer prompts against ChatGPT, Gemini, Claude, Perplexity and Copilot — every answer you are in, every answer you are not, and who was named instead.
Your own Search Console data matched to each gap, so the build order is set by impressions and position rather than by instinct.
BUILD, OPTIMISE or WATCH on every row. You start Monday knowing exactly which page to write first.
Every BUILD row becomes a draft with the FAQ block and the schema attached — see publishing from ChatGPT for how it ships.
Same prompts, same engines. You find out whether the answer changed, which is the only result that counts.
Plans start at USD 39 per month, and the first run costs nothing. Compare that to one afternoon spent arguing about what to write next. See pricing.
Re-run the same prompt set four to six weeks after publishing and compare coverage on the specific prompts you targeted. That is the metric. Site-wide coverage moves too slowly to manage against, and traffic moves for reasons that have nothing to do with you.
If the targeted prompts have not moved after two pages and six weeks, the constraint is not your content — it is authority, and the answer becomes citations and third-party mentions instead of more pages. Knowing which of those two problems you have is worth more than another quarter of publishing.
Create a workspace, track your prompts, and you will have the gap list in about a minute. The build order comes free with it.
Explore next