How we measure AI visibility
We measure observable provider outputs. We do not claim access to hidden ranking logic, secret model weights or a permanent universal score.
The measurement unit is one provider-question observation
For each buyer-style question, the system asks an AI provider for an answer and records whether the target business was surfaced. OpenAI and Gemini are treated separately because they can disagree.
Available now: Free Scan and A$29 Snapshot
The free preview runs three buyer-style questions across OpenAI and Gemini. The current A$29 Snapshot runs 10 representative questions across the two providers, creating 20 provider-question observations.
The paid score is the percentage of those 20 observations in which the target business was surfaced. Competitor count and approximate position do not secretly alter that score.
Snapshot is the existing Full AI Visibility Report shown in the purchase flow. The deeper Pro Audit is a separate request-only scope; it is not an automatic upgrade in that checkout.
Evidence boundaries
We make the report useful without pretending an observed outcome proves causation.
- We can say: what was asked, which provider answered, whether the target appeared, which alternatives appeared and what evidence was observable.
- We cannot honestly say: the exact hidden reason a model omitted a business, the model weight assigned to any signal or that one change guarantees future recommendation placement.
- Evidence Gap Analysis: compares observable public information and highlights differences worth investigating rather than labelling them as secret ranking factors.
- Provider disagreement: is shown rather than averaged away because one provider may surface a business while another does not.
Our deeper Pro benchmark
The request-only A$249 Pro AI Visibility Audit expands the measurement to 40 representative buyer questions across OpenAI and Gemini for 80 provider-question observations, followed by operator review and a prioritised next-step plan.
We intentionally do not publish the exact question taxonomy, provider prompts, source-correlation rules, internal weighting, competitor-analysis heuristics or intervention sequence. Those are developed through our own R&D and retained as service know-how.
Customers receive the measurement record, reviewed findings, evidence boundaries and useful next steps without receiving a copyable recipe for our internal analysis engine.
We test the system on ourselves
Our historical self-audit baseline recorded no measured visibility in the original benchmark. In a later clean comparison, OpenAI surfaced Am I Recommended? in four of five Discovery & shortlist questions at positions #1 through #4.
That movement is useful proof that recommendation visibility can change. It is not proof of one magic source or one guaranteed tactic, and we keep the detailed source correlations and intervention history private while the research matures.
How findings become improvement work
A weak result becomes a reviewed evidence question. We identify the strongest public gaps worth investigating, agree any public changes before implementation, preserve what changed, and later compare results again where a retest is appropriate.
Limitations
- AI answers are probabilistic and can vary across time, model versions and provider settings.
- A point-in-time audit cannot prove market share or total customer demand.
- Public-web evidence can change after the report is generated.
- Visibility is not the same as business quality, legal compliance, customer satisfaction or guaranteed sales.
- A retrieved source appearing in a successful observation does not prove that source caused the model to surface the business.