SpyderBot Documentation
Metric questions and answers
“AI Visibility” is the broad concept. For the Organic Visibility percentage, SpyderBot checks whether your brand is substantively mentioned in usable answers to eligible open discovery questions. It calculates rates within the declared prompt and system configurations, then combines them with the standard balanced method. Core Prompts and independent systems receive equal weight within that profile; collecting more repeats does not give one of them extra influence.
How is AI Visibility calculated?
For a simple complete two-system example, 8 mentions in 10 answers from one system and 2 in 2 from another give rates of 80% and 100%, and balanced visibility of 90%. A full report also balances its declared prompt scope. The raw total, 10/12, is a different pooled statistic. Missing collection is not counted as a brand absence.
Visibility measures representation in the tracked scope. It does not combine sentiment, recommendations, citations or traffic into one score, and it is not the percentage of all AI users who see your brand.
Organic Visibility: definition and calculation ↗
Why does my percentage not match mentions divided by responses?
The headline uses balanced aggregation; raw counts describe collection volume. A system with more accepted repeats must not automatically dominate the headline. Read the prompt and system breakdown and the declared profile.
How balanced aggregation works ↗
Does mentioning my brand mean recommending it?
No. Mention means substantive representation. Candidate Inclusion means an admissible option. Recommendation means active advice to consider, choose or use the entity. Primary Recommendation requires an explicit preference or preferred top tier. Compare these only on the same eligible scope.
Why is AI Search exposure lower than presence within AI answers?
Exposure uses all verified eligible search opportunities, including searches that produce no AI answer. Presence within AI answers uses only generated usable AI answers. If 40 of 100 verified opportunities produce answers and 10 mention your brand, exposure is 10% and presence within answers is 25%.
Does 50/50 global weighting mean half of AI users use each channel?
No. It is the named balanced comparison profile, not estimated audience share. It standardises the relative influence of assistant and search observations in that result.
Why can Sentiment Index be the same for very different answers?
The index compresses a four-label distribution. Neutral and mixed receive midpoint credit, but neutral means no substantive positive or negative tone while mixed contains both. Read the four-label breakdown and classification coverage, not only the index.
Does High evidence mean the answer is true or 95% certain?
No. Evidence level summarises support for the declared measurement. It is separate from claim verification and from a statistical confidence interval. Repeated agreement can still repeat an error.
Why is a metric unavailable rather than zero?
Zero is an observed result. Unavailable means the necessary evidence is missing or unresolved. Not applicable means a required condition is known not to exist. The result’s reason tells you which case applies.
Can I reproduce the number from my exported answers?
Only if the export includes the complete eligible observations, accepted sample selection, scope and aggregation information needed for that result. A flat or partial list of answers is not enough to reconstruct a balanced multi-level headline. Do not infer hidden weights from rounded dashboard values. Use the declared method and the evidence available to your account.
Does a citation mean the source influenced or trained the model?
A citation shows an explicit reference in the observed answer. It does not establish training use or causal influence. Whether SpyderBot could read the source fully is a separate access status.
Why do Quick Overview visits differ from AI Traffic?
Quick Overview uses named external estimates in a frozen snapshot. AI Traffic uses connected observed activity and its own measurement source. They must stay labelled separately; one must not silently replace the other.
Estimated visits ↗ · Observed AI referrals ↗
Did my content change cause the visibility increase?
The measurements show an observed change. A comparable before-and-after result alone does not prove causation: AI-system changes, source changes and ordinary response variation can also contribute. Use the evidence and comparison context to evaluate hypotheses.
What is private about the calculation?
The measurement definition, eligible population, aggregation basis and limitations are explained in the methodology. Internal extraction prompts, recognition and classification implementation, calibration and infrastructure remain private. If you question a result, review the available supporting observations and report the specific metric, period and example through your normal support channel.