Unprompted brand rate
The share of successful discovery answers that mention the brand without the prompt naming it. This is a GeckoCheck measurement convention.
Illustrative calculation, not measured data
- Answers mentioning the brand
- 12
- Successful discovery answers
- 30
12 / 30 × 100 = 40%
What it means for GEO
Separate category discovery from branded verification. Preserve successful-answer counts and log failures separately so the denominator is visible.
What not to infer
A mention does not prove an owned-site citation, positive recommendation or sale.
How it works
This metric asks whether a brand appears in category-discovery answers without being named by the prompt. Decide what counts as a brand mention before reviewing the cohort: a known spelling variant may count, while an unrelated business with the same name should not. Count each qualifying answer once even when the name appears repeatedly. Branded questions, instructions that reveal the target brand and conversations already discussing it belong in a different cohort. The measure describes observed outputs, not the share of all people discovering the brand.
GEO example
Illustrative scenario, not observed assistant behavior or measured results.
Fictional worked example: 30 successful answers to unbranded category prompts include the target brand in 12 answers, giving 40%. If five of those mentions are negative, they still count as mentions under this definition; recommendation and sentiment need separate labels. Suppose two additional requests failed: report 30 successful answers out of 32 attempts, with two failures, rather than silently discarding them or treating them as valid no-mention answers. The graph above uses only the 30 successful answers.
What you can do
- Freeze an unbranded prompt set and document allowed aliases and disambiguation rules.
- Start comparable conversations and record whether memory, history or location could introduce the brand.
- Label recommendations, sentiment and owned citations separately from simple mentions.
What to measure, and what it cannot prove
Report the numerator, successful-answer denominator, total attempts, failures, assistant, mode, language and dates. A higher mention rate after a content change is an association, not proof that the edit caused it. Repeated observations on fixed prompts are not a representative survey of users and should not be described as market share.
Evidence and authoritative sources
- Aggarwal et al.: GEO (KDD 2024)
Motivates explicit, purpose-specific visibility metrics. It does not define or validate this GeckoCheck brand-rate formula.
- OpenAI: Searching the web with ChatGPT
Documents that memory and location can influence search, supporting the need to record observation conditions.
The sources establish the documented concepts and product behavior. The publishing checklists, scenarios and operational measurement rules are GeckoCheck educational interpretations, not provider-endorsed ranking instructions.
Source: Aggarwal et al.: GEO (KDD 2024) · OpenAI: Searching the web with ChatGPT
Period: Reviewed October 12, 2026. Living provider documentation; cited research published in 2023 and 2024 where applicable.
Sample: Fictional example: 30 successful answers, 12 brand mentions. Not collected GeckoCheck or assistant data.
Limitation: GeckoCheck educational interpretation. Examples are fictional; formulas use declared operational definitions, not universal provider metrics. Source documentation does not guarantee citation or ranking outcomes.