THE THING WE BUILT:

Answer Track

A few weeks ago we shared AI Search Readiness. The tool answers one question: does your website have the content to support the buyer conversations you want to win? It doesn’t try to predict ChatGPT rankings or boil everything down to another GEO score. It evaluates your content against real questions, surfaces the conversations your site already supports, and points to the ones you’re still missing.

We got some great responses to that experiment. We used it on ourselves and even tested it with some clients. Then the obvious next question came up: okay, but what do the chat tools actually say today?

There are already plenty of products that monitor AI prompts, from dedicated tools like Otterly, Peec, and Profound to much larger marketing platforms like HubSpot and Semrush. They’re built to do a lot: visibility scores, competitor tracking, citation analysis, recommendations, dashboards, and broader toolsets.

We wanted something simple, but also something that would let us open up the hood and better understand this still-new way people are finding things on the web. No proprietary score to decipher or analytics layer obscuring what happened. Just a simple tool you can understand immediately, with the full responses, logs, and exports there when you want to dig into the process.

Ask the questions you care about. See exactly what the models said. Save the answers. Come back later and see what changed.

So we built a thing.

Answer Track in action

How It Works

Start with the buyer questions you actually care about winning, then run them across ChatGPT, Gemini, Claude and Perplexity.

Each prompt is a fresh ask. No chat history bleeds from one question into the next, so next month’s run is comparable to this one.

Answer Track saves each run so you can compare the results over time: Did we start appearing? Did we disappear? Did our position move? Who are the models recommending instead?

Publish something new, run the same prompts again, and compare the evidence instead of relying on “I asked ChatGPT once.”

Notes from Building

1. The cross-model view is the snapshot. History is the product.

It’s easy enough to ask one chatbot one question. The useful artifact is a dated run across several models, followed by another run after something changed. That is when “not mentioned” becomes “mentioned,” or a buried recommendation moves toward the top.

2. Readiness and Track close the loop.

AI Search Readiness asks: do we have the content to support this buyer conversation? Answer Track asks: are AI tools actually mentioning us for it? Readiness tells you what to write. This gives you a baseline, preserves the answers, and lets you see whether anything changes after you publish it.

3. Keep the evidence; skip the theater.

Other products offer monitoring dashboards, visibility scores, and competitor panels. Answer Track keeps it simpler: ask the same questions across multiple AI models, save exactly what they said, and see what changes over time.

No percentile, letter grade, or “authority index” pretending to measure more than we can actually observe.

The answers are the evidence. The history is what makes them useful.

Built with ❤️ by Township

You’re receiving this because you signed up for Township’s “We Built a Thing” newsletter. We'll only send these when we, well, build things.