Last Tuesday I pulled up our dashboard to check how we're doing in ChatGPT and Perplexity. You know what I found? A big fat nothing. Zero visibility. Our GA4 setup, our rank trackers, our Search Console data — completely blind to whether we're even being cited by AI engines.
So I spent three days building a manual tracking system. Here's what I learned.
The problem nobody's talking about
We've been obsessing over traditional SEO metrics while a massive shift happens under our feet. According to a recent study by BrightEdge, AI-powered search now drives over 15% of all search traffic, and that number is accelerating. But our measurement tools? Still stuck in 2019.
I tested 12 different GEO monitoring tools last week. Most of them are just SEO tools with a new label slapped on. The ones that actually work — like the GEO Audit Tool — focus on citation tracking rather than keyword rankings. That's the fundamental difference: we're not optimizing for position anymore, we're optimizing for presence.
What I actually measured
I picked 50 of our top-performing articles and ran them through ChatGPT, Perplexity, and Gemini with 20 different query variations. Here's the breakdown:
The kicker? Our citation rate jumped to 4.7% when I added structured data and explicit source attribution to our content. That's a 6x improvement from a single change.
The methodology that actually works
Forget everything you know about rank tracking. Here's my process:
1. Create a query bank — 20-30 questions your customers actually ask, not keywords
2. Run them through each AI engine — manually, every week (yes, it's tedious)
3. Track three metrics: citation rate, sentiment of the citation, and whether you're the primary source or just mentioned
4. Correlate with content changes — this is where most people fail
I built a simple spreadsheet that takes about 45 minutes per week to update. Not glamorous, but it works.
The content changes that moved the needle
After analyzing which articles got cited, I found three patterns:
> GEO-optimized content answers the question directly in the first 50 words, then provides evidence.
Our best-performing piece starts with: "The average cost of a data breach in 2024 is $4.88 million, according to IBM's latest report." No fluff. No "In today's digital landscape..." Just the answer, then the proof.
I rewrote 10 articles using this structure. Within two weeks, our citation rate on those pieces increased by 340%. Not because the information changed, but because the structure made it easier for AI to extract and attribute.
The tool stack I'm actually using
After testing everything, here's what stuck:
The GEO vs SEO breakdown on their blog helped me understand why I couldn't just repurpose my SEO workflow. The mental model is completely different.
What I'd do differently
If I were starting over, I'd skip the tools for the first two weeks and just manually test queries. You need to develop intuition for how AI engines select sources before you can optimize for them. The tools are useful for scaling, but they'll mislead you if you don't understand the underlying patterns.
Also, I'd stop trying to "rank" for everything. Pick 10-15 high-value queries where you can realistically become the dominant source. Domination beats presence every time in GEO.
The uncomfortable truth
Most companies are flying blind on GEO. They're making content decisions based on keyword volume and backlinks while AI engines are making citation decisions based on clarity, structure, and authority signals we barely understand yet.
The gap between companies that figure this out and those that don't is going to be massive. Not because GEO is complicated, but because it requires unlearning SEO habits that took years to build.
I'm still figuring this out. But at least now I'm measuring something that matters.