The words AI associates with your brand

What is Semantic Perception?

Semantic Perception is the linguistic analysis of every word AI models attach to your brand in their answers. Where Source Authority looks at the URLs cited and the audit scores look at frequency, Semantic Perception looks at the texture of what is said.

It tells you whether ChatGPT calls you “innovative” or “slow”, “premium” or “expensive”, “secure” or “complex” — and how often.

How the analysis works

1. Response collection. Each prompt run during the audit produces an AI answer. The full text is stored, with the model and market it came from.
2. Linguistic extraction. A two-pass extractor identifies the adjectives, verbs and noun phrases attached to your brand name in the same sentence (or the directly preceding one). Stop-words and generic filler are removed.
3. Sentiment scoring. Each extracted token is scored positive / neutral / negative using a multilingual lexicon plus an LLM verification pass on ambiguous items (“cheap” can be praise or critique depending on context).
4. Comparison with competitors. The same extraction runs on competitor mentions so you see whether the negative term “complex” sticks to you specifically or to the whole category.
5. Multilingual aggregation. Words are normalised per language but kept distinct per market — what is praise in one country can be neutral in another.

Reading the dashboard

In the Dashboard v4, Semantic Perception sits in Zone 2 — Presence in AI answers, just below Source Authority. The hero element is a coloured word cloud where the size of each word reflects its frequency, and the colour reflects its sentiment.

The panel ships with three companion views:

• Top 15 ranking — the same words as a sortable list with absolute frequencies.
• Sentiment breakdown — a stacked bar showing the share of positive vs neutral vs negative on the brand. Useful for executive reporting.
• Per-market matrix — your top 5 terms for each language, side by side.

Tracking evolution

With at least two audits in the history, the panel shows a delta column: which words are new, which gained traction, which dropped. A green up-arrow next to a positive term is a win. A green up-arrow next to a negative term is a warning.

When you mark a term as “watched”, it gets pinned at the top of the list across every future audit so you can read its trajectory week after week.

Typical use case

A B2B SaaS sells the same product across France and Germany. Semantic Perception reveals two very different brand portraits:

Same product, same prompts, two opposite narratives. The action is concrete: rework the French landing pages and customer stories to lead with simplicity and ROI, keep the German messaging that is already working.

Plan and access

Semantic Perception requires the Consultant+ plan or higher. The Agent and Agence+ plans add the per-market matrix and unlock exporting the word cloud as a PNG into your white-label PDF.

Going further

• Blog: AI semantic perception — what words AI associates with your brand
• Source Authority tutorial
• Dashboard tutorial — 5-zone architecture