rung 3 of 5 · the specimen page

Referenced, are you the source answers are built from?

The rung where AI search stops resembling classical SEO. You can rank well and be cited never, and the two are measured on different surfaces.

What is measured

Citation rate and cited URLs per answer engine, share of voice against tracked rivals, and prompt coverage gaps, kept strictly separate from Google's on-SERP AI Overviews, which is a different surface.

What it is routinely confused with

Four distinctions do most of the work on this rung. Each is a real reporting error, not a pedantic one, every one of them changes what you would do next.

• Answer-engine citations are not AI Overviews

  A citation in a ChatGPT or Perplexity answer and an inclusion in Google's on-SERP AI Overview are different surfaces, collected differently, moving independently. A single blended "AI visibility" number hides which one moved, and therefore what to do about it.

• Citation is not the same as mention

  Being named in an answer and being linked as the source it was built from are different events. Both are measured; they are reported as separate figures, because a brand can be mentioned constantly and cited never.

• Share of voice here is not search market share

  The AI figure is a share of citations across a tracked prompt set. It carries no CTR model and no click estimate, so it must never inherit the qualifier that belongs to CTR-modeled search market share, nor be compared to it as if the two were the same unit.

• A prompt basket is a sample, not a census

  Every AI figure is scoped to the prompts being tracked. Hand-picked baskets flatter; the coverage of the basket belongs on the card next to the number it produced.

What a reading looks like

A simulated conversation on fictional data. Note the per-engine split: one blended number would have hidden that a single engine carries the whole result.

The levers that move this rung

• secondary L2 · Demand modeling → Which demand is worth entering at all, modelled from your own data rather than a third-party volume estimate.
• secondary L3 · Refresh & content quality → Half of the wedge: already-published pages, so the loop closes in weeks.
• secondary L4 · Internal linking & architecture → The other half of the wedge, and the cheapest lever that moves two rungs at once.
• primary L5 · Net-new content → The most expensive lever and the one where humans stay mandatory. Scaled last, after cheaper levers have proved the demand.
• primary L6 · Authority & off-site → Continuous rather than sequenced, it compounds and it cannot be sprinted.
• primary L7 · AI-answer visibility → Worked through prompt baskets rather than keywords, because the unit of demand is different.

Where it shows up

• AI visibility standing UC-04 · Weekly
• AI crawler readiness UC-05 · Monthly