ELI RESEARCH
AI-search recommendation rate: measure how often a brand is actually recommended
A mention or citation is not automatically a recommendation. Measure whether a brand is an option in the buyer decision.
Benchmark not yet released. Eli has defined methodology and is collecting the first observation set. No industry average appears until release criteria are met.
RELEASE STANDARD
Methodology is publishable before findings are.
Eli can publish the question universe, schema, definitions and quality controls before a benchmark release. It will only publish an aggregate when the underlying observations are complete, reviewable and described with their coverage and limitations. Until then, this page makes no performance, ranking or customer-outcome claim.
Formula
Recommendation rate equals recommended observations divided by eligible observations. Denominator must be explicit; do not add unrelated prompts to change rate.
Recommendation vs mention
A brand appearing as vendor differs from being recommended for the use case. Publish classification rules and flag ambiguity.
Recommendation context
Track first option, one of several, conditional recommendation, negative/avoid context and comparison-only mention.
Future views
When real release exists, show distribution, median, quartiles, buyer-stage, engine and category comparisons only for groups meeting thresholds.
How to use
Use rate for questions that matter, repeated improvement, weak intents and competitor wins; not as standalone revenue metric.
FAQ
Questions about the research design and release boundary.
Good rate?
No universal number before relevant benchmark/universe.
Share of voice?
Related, not identical.
Should branded prompts count?
Report branded and non-branded separately.
Mentioned not recommended?
Yes.
Higher means revenue?
Not necessarily.
CONNECT YOUR WEBSITE
Measure your recommendation rate
Recommendation rate is the share of eligible AI-answer observations in which a brand is presented as a relevant recommended option. Measure it across a fixed prompt universe and repeated observations, not one screenshot.
Measure your recommendation rate