Fan-out Queries

Inspect the internal queries LLMs use to build their answers and where your domain fits in.

To build a response, models often generate multiple internal queries — known as fan-out queries — to explore a topic before producing the final answer. These internal queries largely determine which angles of the topic are explored, which domains and URLs are retrieved, and whether your domain or brand later appears as a Source or a Link.

The Fan-out Queries view exposes this internal exploration so you can understand how generative answers are actually constructed.

Read What is query fan-out and why it matters in LLM search for the conceptual primer.

What this view helps you analyse

The view bridges LLM reasoning and search visibility signals by showing:

• how LLMs decompose a prompt into multiple internal queries,
• how stable or volatile these fan-out queries are,
• how engines differ in fan-out behaviour,

Identify global fan-out patterns

Fan-out overview dashboard

The Overview surfaces high-level patterns across fan-out queries:

• Total QFO — total number of fan-out queries generated across tracked prompts
• Average QFO — average number of fan-out queries per prompt, by engine

You can also break coverage down by thematic / topic and by intent type (exploratory, comparative, decision-making, etc.).

This view answers questions like:

• How broad is the fan-out explored by the model?
• On which topics or intent types is my coverage stronger or weaker?
• Do different engines behave differently at a high level?

Analyse individual fan-out queries

Detailed fan-out queries list

Scroll down to the detailed list, displayed either flat (Grouped by: default) or grouped by prompt for cross-engine comparison. For each fan-out query you see:

• the fan-out query itself
• its associated intents
• its stability over the last 12 collects (one bar per collect where the fan-out was present)

You can:

• understand which internal queries the model relies on repeatedly,
• distinguish structural fan-out queries from occasional ones.

Depending on the model and API capabilities, fan-out visibility may be partial. It reflects the retrievable part of the model’s exploration, not its full internal reasoning.

Understand how a single prompt is decomposed

Prompt-level fan-out

From the Rankings or Prompts views you can drill into fan-out details for a specific prompt. This shows how a single prompt is broken down, and which fan-out types the model prioritises.

Reading cues

These are signals to read, not rules.

• Fan-out queries that are stable over time often reflect core questions the model consistently asks itself.
• Differences across engines highlight engine-specific reasoning strategies.
• Recurring fan-out queries can be read as indicators of topics the model expects to find content about.

Fan-out data should be read as a map of the model’s internal exploration — not a traditional ranking report.

How to use this view effectively

Start by identifying fan-out queries that appear consistently across engines and collects. These recurring queries reflect the core questions the model relies on to determine which domains influence the final answer.

What’s next

Query Fan-out Explained

Conceptual primer on fan-out behaviour.

Sources & Links

See how fan-out outcomes turn into sources and links.