Skip to main content

​
Direct answer

Use keywords as research labels, not as a promise of inclusion in AI answers. Combine configured prompts, observed Query Fan-Out records, citation-source language, and Google Search Console data to identify audience needs worth testing with content.

​
Data required

Use your configured prompts, topics and tags, stored answers, observed Query Fan-Out records, Source Explorer, and connected Google Search Console queries. Keep GSC search queries distinct from Qwairy prompts and from provider-generated web queries.

​
Workflow

1

Collect candidate language

Export relevant prompts, observed web queries, source titles, and GSC queries for a defined topic and period.
2

Group by audience need

Cluster terms by the question or decision they represent. Use topics as the primary semantic grouping and tags as reusable secondary labels.
3

Check current evidence

Read the stored answers and cited sources for each group. Note where your brand or first-party sources are present, absent, or represented inaccurately.
4

Prioritize tests

Select terms that are relevant to the business and supported by observed demand or coverage gaps. Define the content change and the metric slice you will monitor.

​
Interpretation

A keyword list is a planning artifact. It does not show the exact input a model used or guarantee a citation. Evaluate changes on stable prompt and provider cohorts. Product Mention Rate is answers with a SELF mention divided by answers with at least one SELF or DIRECT brand mention; it is not a keyword ranking metric.

​
Possible next actions

  • If GSC demand is present but SELF citations are limited, test content that answers the same audience need with verifiable evidence.
  • If Query Fan-Out reveals recurring observed searches, consider adding a representative configured prompt before creating content.
  • If one topic contains unrelated intents, split the topic and keep cross-cutting labels as tags.

​
Limitations

Query Fan-Out is an observed subset, not an exhaustive list of what people or models search for. GSC and Qwairy measure different surfaces. Correlation between a term, a ranking, and an AI citation does not establish causation.

Exports

Export the selected dataset for research.

Keyword Performance

Review the Looker Studio source for keyword reporting.

Google Search Console

Understand the connected traditional-search data.

Choose what to create

Turn a research list into content hypotheses.

Google rank versus AI presence

Compare two distinct measurement surfaces.

Gap to article

Run a documented publishing experiment.

Pages cited by AI

Inspect the URLs recorded as citations.