Recommendation share

Recommendation share is a cleaner way to talk about AI visibility because it asks how often a brand is actually recommended for the questions that matter.

Definition

Recommendation share is the share of a stable buyer-question set where an AI answer recommends a brand with meaningful strength. It is different from mention share. A brand can be mentioned often without being framed as a good fit. Recommendation share should be measured against a defined denominator: which prompt families, answer surfaces, markets, competitors, and scoring threshold are included.

Why marketers care

Teams need a metric, but the wrong metric can flatter them. Mention share counts appearances. Recommendation share asks whether the answer gives the buyer a reason to consider the brand. That makes it more useful for prioritizing positioning, proof, comparison, and source work.

The denominator matters

Recommendation share is only meaningful when the question set is defined. Ten friendly branded prompts will produce a very different number from fifty non-branded category, comparison, and objection prompts. Always name the prompt families, surfaces, market, and cadence.

What counts as a recommendation

A practical threshold should include more than a bare mention. The answer should create buyer confidence or fit.

  • The brand is recommended for the stated use case.
  • The answer gives a reason, proof point, or fit statement.
  • The brand is placed in a shortlist with useful context.
  • The recommendation is not undermined by a stronger caveat than the competitors receive.

How to calculate it

Define the question set, read the answers, classify recommendation strength, then calculate recommended answers divided by eligible answers. Track the result by prompt family and surface instead of collapsing everything into one vanity number.

A practical example

If a brand is recommended in 9 of 30 high-intent questions, its recommendation share is 30 percent for that set. If it is mentioned in 20 of 30 but recommended in only 9, the gap shows that presence is not the problem. Persuasion is.

Common confusion

Recommendation share is not market share, search share, or share of voice. It is an answer-layer metric for a controlled read. It should guide work, not pretend to measure total demand.

Signalbat interpretation

Signalbat treats recommendation share as one useful rollup over richer observations. The Reading still needs to explain where the share changed and what source, claim, competitor, or proof gap caused the movement.

How to measure it

  1. 01

    Define the set

    Lock the prompt families, answer surfaces, market, and competitors included in the read.

  2. 02

    Score strength

    Classify each answer by recommendation strength, not just brand presence.

  3. 03

    Explain movement

    Break changes down by question family so the number turns into work.

Inputs behind the metric

  • Question denominator
  • Prompt family
  • Surface coverage
  • Strength threshold
  • Eligible answers
  • Recommended answers
  • Competitor recommendations
  • Change from prior read

Recommendation share questions

Is recommendation share better than mention share?
Usually, yes. Mention share can be useful as a presence signal, but recommendation share is closer to what a buyer hears when deciding which brand deserves attention.
Can recommendation share be compared across companies?
Only if the same question set, surfaces, markets, cadence, and scoring rules are used. Otherwise the comparison is mostly noise.
What should a team do when recommendation share drops?
Inspect the questions that moved. Look for new competitors, weaker wording, caveats, source changes, or proof gaps before deciding what to fix.

Research behind this definition