Skip to content
NewNew: Autopilot Agents find competitor gaps while you sleep.Read the note →
SearchDockSearchDock

How SearchDock measures AI search visibility

SearchDock measures whether AI answer engines mention and cite your brand, and how that compares with competitors and with your Google rankings. This page explains exactly how those numbers are produced so you can judge what they mean · and where a figure is observed capture data versus a directional estimate.

Engines we sample

SearchDock samples AI answers across ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews, alongside classic Google organic rankings pulled from your connected Google Search Console property. Coverage of a given engine depends on your plan and on API availability at the time of capture · a page notes when an engine was not sampled rather than reporting a zero.

Multi-pass prompt sampling

AI answers vary from run to run. A single query can name your brand once and omit it the next time. To avoid reporting one lucky (or unlucky) response, SearchDock runs every tracked prompt multiple times per engine and reports the average across those passes · so a citation rate reflects how often you appear, not whether you appeared once.

  • Prompts are grouped by product, category, and buyer intent so rates are comparable over time.
  • Rates are computed from the stored capture history · not from a single live call at page load.
  • New prompts need several passes per engine before a stable average is shown.

How we define a mention and a citation

These two are measured separately, because they answer different questions:

  • Brand mention · the engine names your brand or product in its answer text. This is a raw, observed count.
  • Owned citation · the engine links to a URL on your own domain (a domain match). "Earned" references point to third-party pages that mention you.

Raw mention and citation counts are real, observed data. We never infer a mention that did not appear in a captured answer.

Share of voice

Share of voice is computed from stored captures across the same prompt set for you and the competitors you have configured · not from an open-ended guess about "everyone" in your market. A competitor only enters the calculation once you add its domain, so the denominator is always a set you defined.

Observed data versus directional estimates

SearchDock labels these differently, and so should you when you quote them:

  • Observed · mention counts, citation counts, and share-of-voice figures computed from real captured answers.
  • Directional · opportunity scores, forecasts, and scenario simulations. These are estimates meant to prioritize work, not measurements · they are marked as directional wherever they appear.

Dashboards read from stored capture results · they do not call AI models live when a page loads. That keeps figures stable between views and makes them reproducible.

Data freshness and variability

Captures are refreshed on a schedule, and each result carries the time it was measured. AI answers also vary by region, personalization, and model version · a rate is representative of the prompts, engines, and window sampled, not a guarantee for every user in every location. Where a page shows a rate, it also shows the sample it came from.

What we do not do

  • We do not fabricate rankings, citations, competitors, or results · missing data is shown as an honest empty state with a path to fix it.
  • We do not invent ratings, reviews, or prices in generated schema or content.
  • We do not present directional estimates as observed measurements.

Questions about how a specific metric is produced? SeeAEO tracking,AI citations, andAI share of voice, orcontact us.