AEO Tracker
What four LLMs say when you ask them about your brand.
I built a tool that runs the same prompts across ChatGPT, Claude, Gemini, and Perplexity, then logs how each one talks about a given brand. I was trying to build to something I hadn't seen in the market yet, what I was calling "message match." The idea was that an LLM was a reflection of how you were being perceived by the world because it's trained on all sorts of different training data. I wanted to add to the existing LLM visibility tools a function of "Is the LLM saying what I want it to about me?"
What surprised me
- The "one prompt five minutes working app" lines are mostly hype. I could've done better, but the method I took required a ton of babysitting.
- Related... The "GSD" extension is high-powered. I think overpowered for solo work like this. I ended up with over 44 phases and a pile of jumbled mess at the end.
- That pile of jumbled mess was mostly because I let my front-end and backend drift. I'd built the prototype as a CLI tool, then I tried to add a UI on top without enforcing contracts. testing became a mess and almost seemed like the entire website was hallucinating at me. The front end would say it was going to do something that the backend had no idea about, etc
What it doesn’t do
- Doesn’t track citations.
- Doesn’t tell you why.
- Doesn’t optimize anything.
Ultimately, I found a solution that was close enough to what I needed on the market so abandoned this project having learned quite a bit.