Trust and measurement

Methodology

See exactly what Crawlbird measures, how every percentage is calculated, and where the benchmark has limits.

What this benchmark means

Answers captured from ChatGPT. Results describe the tested sessions; personal answers may differ. Failed captures are excluded and API answers are never substituted. This measures the approved buyer questions, not search volume or purchase probability.

Independent runs

3 per prompt

Model

Reported per answer

Market

Project setting

Web search

Requested; use shown per answer

How the metrics are calculated

Presence Rate: Valid runs where the brand appears ÷ total valid runs × 100.
Recommendation Rate: Valid runs where the brand is genuinely recommended or shortlisted ÷ total valid runs × 100.
Discovery Visibility: Weighted average Recommendation Rate across approved non-branded commercial prompts. Low importance weighs 1×, medium 1.5× and high 2×. Branded checks are excluded.
Share of Voice: Brand recommendation appearances ÷ recommendation appearances for every tracked brand, within this prompt set.
Competitor Gap: Top competitor Recommendation Rate minus your Recommendation Rate, measured over the same tracked prompt set.
Citation Rate: Valid runs that cite the brand’s domain ÷ total valid runs × 100. This is separate from recommendations.
Run agreement: How many independent runs agree on whether a brand was recommended. Three of three is more stable than two of three, but neither is certainty.
Average position: Mean position only across answers with a meaningful explicit ordered shortlist. Paragraph and bullet order never create a rank.

Why the ChatGPT app may answer differently

Personal ChatGPT answers can differ because of model variation, location, account history and personalisation. This report measures a consistent, repeatable baseline so that changes can be compared over time. It cannot prove that a cited page caused a recommendation, how many people ask a prompt, or that a content change will improve the result.

Language: en · Last data update: No completed scan yet

Delete project data

Permanently removes the project, tracked prompts, every raw OpenAI response, citations, metrics and usage history.

How the metrics are calculated

Presence Rate

Valid runs where the brand appears ÷ total valid runs × 100.

Recommendation Rate

Valid runs where the brand is genuinely recommended or shortlisted ÷ total valid runs × 100.

Discovery Visibility

Weighted average Recommendation Rate across approved non-branded commercial prompts. Low importance weighs 1×, medium 1.5× and high 2×. Branded checks are excluded.

Share of Voice

Brand recommendation appearances ÷ recommendation appearances for every tracked brand, within this prompt set.

Competitor Gap

Top competitor Recommendation Rate minus your Recommendation Rate, measured over the same tracked prompt set.

Citation Rate

Valid runs that cite the brand’s domain ÷ total valid runs × 100. This is separate from recommendations.

Run agreement

How many independent runs agree on whether a brand was recommended. Three of three is more stable than two of three, but neither is certainty.

Average position

Mean position only across answers with a meaningful explicit ordered shortlist. Paragraph and bullet order never create a rank.

Why the ChatGPT app may answer differently