Feb 5, 2026

Part 5: Running Your First GEO Audit (The Hard Way)

You have your personas.
You have your topics.
You have hundreds of queries you want to understand.

Now what?

You run them.

In the last post, I covered how query types and intent stages multiply into hundreds of queries. Today, I want to walk through what it actually looks like to run a GEO audit manually.

This is the hard way. But it is worth doing once.

Start small

Do not start with 300 queries.

Pick 20.

Make sure they span your three query types:

Branded
Non branded
Competitive

Open ChatGPT and run each query one by one.

What to track for each query

For every response, capture the same four things.

Mentions
Did your brand appear at all? Yes or no.

Sentiment
How was your brand framed? Positive, neutral, or negative.

Competitors
Which other brands appeared in the answer? Track all of them.

Sources
Which websites did the model cite or reference? These are the authorities AI trusts in your category.

Create a spreadsheet. One row per query. One column per field. Run your 20 queries and fill it in.

If you stay focused, this takes about an hour.

What you start to see

Once the data is in front of you, patterns emerge quickly.

Maybe your brand shows up reliably for branded queries but disappears for non branded discovery queries. Maybe one competitor dominates commercial intent questions. Maybe the same three websites get cited over and over again.

This is real signal. This is what AI systems are actually doing today when buyers ask questions in your category.