Read Reports and Responses

Start with run status

Confirm that the run and its model tasks completed. Failed, skipped, or unsupported tasks affect the sample size and should not be treated as valid negative answers.

Overview metrics

Mention rate

The share of valid answers in which the brand was identified. The denominator matters, especially in small samples.

Recommendation rate

The share of valid answers that present the brand as a suitable option. Recommendation is stronger than a mention.

Position

The first location of the brand in the answer or its list order when applicable. It is supporting evidence, not a universal AI rank.

Citation rate and sources

When a model route exposes sources, BrandGeo records URLs and domains. Citation states distinguish available data from unsupported or temporarily missing data. An unavailable citation is not the same as a confirmed count of zero.

Read prompt-level evidence

Expand a prompt row to review:

• the complete model answer;
• whether the brand was mentioned or recommended;
• other brands found in the response;
• available citations and source domains;
• the model and prompt used for the sample.

The exact answer is the source of truth for interpreting a metric.

Prioritized problems

Report recommendations are derived from observed gaps, such as:

• missing from non-branded discovery questions;
• mentioned without being recommended;
• competitors repeatedly occupying recommendations;
• official website content absent from exposed citations.

Use these as investigation priorities, not automatic instructions to publish unsupported content.

Compare history

Use equivalent runs when reviewing trends. Check prompt edits, market changes, model-route updates, and sample size before attributing movement to a content change.

Paid plans may provide CSV export. Exported data supports offline analysis and client reporting, but the same measurement boundaries still apply.