Tree Walker
Our brand-analysis tool that reveals how Gemini talks about a brand by walking the probabilistic paths of the model's own language.
Artificial intelligence models often feel like black boxes, making it hard to know how they actually perceive your brand. A new analysis tool called Tree Walker changes that by showing you exactly how Google’s Gemini understands and describes your business.
When you give Tree Walker your website, it uses Gemini to generate a description of your brand. But it does not stop there. It looks deep into the model’s vocabulary, focusing on two key signals: word rarity and word uncertainty. Word uncertainty is especially important. It reveals the exact moments where the model hesitated before choosing a word. These high-uncertainty words are the weak spots in your brand's AI profile, making them your highest-priority targets for new content.
The tool also maps out the alternative paths the language could have taken. By watching for what is known as token surprise, it builds dozens of viable sentences the model might have produced instead, and surfaces the runner-up words behind every choice.
Together, these features move AI brand visibility out of the realm of guesswork. By reading the model’s own token probabilities, you can see exactly where your brand's story is fragile, and gain a clear, directable process for shaping how AI represents you to the world.
Tree Walker is our analysis tool for seeing how an AI model like Google's Gemini understands and describes a brand. You give it a website, and it generates an initial description of the brand through Gemini's eyes, then explores the probabilistic paths that language could have taken — turning the "black box" of AI perception into something you can measure and direct.
Two signals sit at the centre of the initial analysis: word rarity (how common or rare each term is within the model's vocabulary) and word uncertainty (where the model hesitated before choosing a word). Words flagged as high-uncertainty are the model's weak spots about your brand, and your highest-priority targets for content work.
The core Probability Tree Walker algorithm follows the reasonable alternative paths a sentence could have taken, watching for "token surprise," to build dozens of viable sentences the model might have produced. An Alternative Token Explorer surfaces the runner-up words behind each choice. Together they move brand AI visibility from guesswork to a directable process, reading the model's own token probabilities to expose where a brand's story is fragile.
Link: https://treewalker.ai/