Skip to main content
Every figure in Narrative comes from one pipeline. It starts with what your brand sells and who buys it, finds the searches those buyers make, groups them into topics by what Google itself groups them by, and then connects the prompts people ask ChatGPT, Perplexity and Gemini back to those same topics. This section explains each stage in plain terms, so you can read a topic score or a prompt mapping and know what produced it.
Your brand, keywords from eight sources, judged by two gates and one judge, Google results, topics grouped by shared pages, prompts mapped to topics.Your brand, keywords from eight sources, judged by two gates and one judge, Google results, topics grouped by shared pages, prompts mapped to topics.

The pipeline, left to right. Each stage reads what the one before it produced.

​
The stages

1

Your brand

Narrative starts from what it knows about you: your products, value propositions, differentiators, the audiences you sell to, the competitors you confirmed, and a handful of anchor queries, the head terms you expect your buyers to type. You can write the anchors yourself during onboarding; where you leave a slot empty, Narrative derives one from your brand intelligence. An anchor you wrote is never replaced.
2

Keywords

From that starting point Narrative collects thousands of candidate searches per language from eight sources, then judges each one on whether your buyer would type it. The survivors get a monthly volume, a difficulty, an intent read from the results page, and a score. How keywords are found and judged.
3

Google results

For every keyword that passes, Narrative fetches the first page of Google as a searcher in your country sees it, on the device you chose, along with the AI Overview when Google shows one. Those results are the evidence everything downstream is built on.
4

Topics

Keywords whose results share the same pages are grouped into one topic. Google already decided those searches are about the same thing, so Narrative follows that decision rather than guessing from the words. Each topic is named in both of your languages, scored, and attributed to whoever is winning it. How topics are formed.
5

Prompts

The prompts Narrative asks AI assistants on your behalf are linked to the topics they are about. That is what lets a topic carry both a search reading and an AI-assistant reading, and it decides which topics are measured and which are content-only. How prompts map to topics.

​
Three rules that hold throughout

A decision you make outranks anything Narrative produced. Rename a topic, merge two, move a keyword, reject one, add your own, pin a prompt to a topic: the pipeline reads those as inputs and no rebuild undoes them. Nothing automated overwrites a person’s call. Unknown is not zero. When a provider has no volume for a keyword, when no model could judge a page, when no prompt has been asked yet, Narrative leaves the figure blank and says so. A blank in Narrative means nobody has measured it; a zero means somebody did and the answer was none. Every run is repeatable. The same brand with the same inputs produces the same keywords and the same topics. The lists Narrative generates for your brand are kept and reused rather than regenerated on every run, so your topics do not drift between Monday and Friday because a model felt differently. They change when your brand changes, or when the market does.

​
Where you see it

  • Topics in the sidebar: every topic, its demand, who wins it, how many of your prompts are about it, and whether it is measured.
  • Keywords in the sidebar: every keyword, its volume, difficulty, intent, the results it returned and the topic it sits in.
  • Personalize → Demand: the country Narrative searches from, the device, and the split between your languages.
  • Prompts: each prompt with the topics it is mapped to.
Topics and keywords are re-measured on a weekly cadence, so the figures you read are at most a week old.