04 / Case Study
04GEO Tracker
Not just visibility: the earned-media play that wins AI search.
Most GEO tools just measure your visibility. This one tells you what to do about it.
Runs frontier AI models against keyword-variation clusters to measure who’s getting cited for the terms a business wants to win, then surfaces the publications to target so earned-media effort actually moves the needle in AI search.
Problem
GEO dashboards measure. They don’t move the needle.
AI search visibility matters more every quarter. But most GEO/AI-SEO tools stop at measurement: “here’s your share of voice in ChatGPT.” That’s a number, not a strategy.
When a business under-indexes on AI citations, the question isn’t “are we under-indexed?” It’s “what do we do?” Most tools leave that to the customer to figure out.
Strategic Approach
Treat AI search as an earned-media problem.
When third-party publications dominate citations for your keyword cluster, those publications are the leverage. The right move isn’t producing more content; it’s earning mentions in the outlets the AI is already reading.
So GEO Tracker doesn’t stop at measurement. It identifies which publications dominate your target clusters, and turns that into an actionable earned-media target list: who to pitch, who to network with, who to engage. AI search becomes an earned-media problem, not a content-production problem.
How It Works
Cluster, query, measure, target.
Step one: define what to rank for. Step two: a low-cost model generates a cluster of keyword variations, because users phrase the same question many different ways, and frontier models answer differently based on phrasing, history, temperature, and which model they’re using. Step three: run the cluster across frontier models and capture citations + mentions. Step four: identify the third-party publications dominating the cluster.
Step five, the part that matters: turn those publications into an earned-media target list. Track visibility over time as a secondary measurement, not the headline.
Method
The method, end to end.
The software is built for client work, so the tool itself stays private. The method behind it: one seed term fans out into a cluster of phrasings, every phrasing goes to several frontier models, and the answers converge on the handful of sources worth earning a mention from.
Method schematic
01
Define what to win
Name the terms and questions the business actually wants to own in AI search.
02
Expand the cluster
A low-cost model generates a cluster of keyword variations, because people phrase the same question many different ways.
03
Query the frontier models
Every variation runs against several frontier models. Phrasing, history, temperature and model choice all change the answer.
04
Capture citations and mentions
Record who gets cited and mentioned across the whole cluster: you, competitors, third parties.
05
See who owns the cluster
The same third-party publications keep surfacing. Those are the ones holding the citations you want.
06
Turn it into a target list
Those publications become the earned-media list: who to pitch, who to network with. Visibility over time tracks alongside it, as the secondary read.
Outcomes
What it changed.
- Closes the loop most GEO tools leave open: the output isn’t a visibility score, it’s the specific publications and outlets to network with, pitch, and earn mentions from.
- Measuring a cluster of keyword variations rather than one phrase reflects how people actually ask: the same question worded differently pulls different citations from different models.
- Reframes AI search as an earned-media problem: when third parties dominate the citations for your cluster, earning a mention there is the move, not producing more content.