Core Concepts

Trakkr measures how the world's AI models talk about your brand. Most of the product is built on a handful of ideas - once they're loaded, the rest of the docs read in plain English.

This page is the mental model. Read it once.

The 60-second model

A search engine returns ten links and lets you choose. An AI assistant returns one answer and decides for you.

Trakkr's job is to find out whether your brand made it into that answer - on every question that matters to your business, across every model your customers are using.

When AI doesn't name you, you're invisible to a slice of buyers you'll never measure with web analytics. That's the gap Trakkr fills.

The vocabulary

Six nouns appear on nearly every screen. Once you know these, everything else hangs off them.

One process knits them together: a research run asks every prompt to every model and captures every mention and citation that comes back. Trakkr does one of these every day.

Why you appear (or don't)

When someone asks Perplexity for the best CRM, three things decide whether your brand is named.

1. Training data presence

Models learn from the open internet - Wikipedia, news, reviews, well-linked blogs. If your brand showed up in quality sources during training, the model knows you exist. Paid ads, most social posts, and gated content don't make it in.

2. Contextual relevance

Knowing you exist isn't enough. The model has to connect your brand to the right category. "An all-in-one workspace" is harder to place than "a note-taking app for teams." Clarity wins.

3. Sentiment and positioning

How you're described matters as much as whether you're named. If the sources AI trusts position you as a leader, that follows you into the answer. If they position you as the alternative-to-X, that follows too.

These three are what every Trakkr feature ultimately moves. Citations is the most direct lever - change what models read about you, and the answer changes.

The eight models we track

ChatGPT

Claude

Gemini

Perplexity

AI Overviews

Grok

Meta AI

DeepSeek

Each model has different training data, a different cutoff date, and a different way of pulling in fresh information. Scoring 80% on Claude and 30% on Perplexity is normal - and both numbers are useful signals.

There's no single number that summarizes everything - and that's the point. Different models reach different audiences, and an improvement plan usually targets one tier at a time.

The visibility loop

Improving AI visibility isn't a one-time audit. It's a loop, and Trakkr is built around it.

Measure with Research and the Dashboard - establish a baseline. Understand with Citations and Perception - see what's driving the numbers. Act with Content, Optimize, AI Pages, and Actions - change what the model sees. Re-measure the next day. Live-retrieval models (Perplexity, AI Overviews) move fast. Training-data models take longer.

The map - where each idea lives

The rest of the docs are organized by what you want to do. Use this as a routing index.

Going deeper

Three reference pages cover the long-tail detail. They're worth bookmarking, not reading top-to-bottom.