What do we measure?
| Outcome | What counts | What it does not establish |
|---|---|---|
| Business mentioned | An answer names the verified business. | A website citation or a recommendation in every answer. |
| Website cited | The answer explicitly cites a URL on the business’s verified website. | A source that was merely retrieved in the background. |
| Directory cited | An explicit directory URL; a client profile and general directory page are reported separately. | A citation to the business’s own website. |
| Inquiry or booking | A separately recorded customer interaction, with its reported source. | Proof that our edit caused the inquiry. |
Freeze the questions before measuring
We agree the business identity, service, location, questions, engine settings and repetition count first. A typical pilot uses ten questions and three repetitions per selected engine. We call the discipline “same questions twice”: the follow-up repeats the baseline exactly, so a difference means something. Changing the questions, engines or settings requires a new baseline.
Keep the evidence, including failures
We retain full answers, citation URLs, provider information and attempt history. Failed or incomplete measurements are withheld, not reported as zero visibility. We report each engine separately and compare the latest batch in each phase, not whichever batch looks best.
Make one documented change
We choose an intervention from a verified business-specific issue, record the reasoning and get approval. Publication records include what changed, where and when. The workflow limits interventions to one per client per UTC calendar month.
Interpret the follow-up carefully
Answers can change because of model updates, seasonality, news, competitors or other marketing. Three repetitions help reveal variation, but do not establish statistical significance or causality. A before-and-after difference is an observation.
Which search experiences are covered?
The existing integrations cover Claude, OpenAI, Perplexity and Google AI Mode via DataForSEO. The selected engines and settings are confirmed for each pilot. API samples are not identical to personalized consumer chatbot experiences. Google AI Mode is not Google Maps, ordinary organic ranking, AI Overviews or the consumer Gemini application.
We apply the same standard to ourselves
Found & Proven’s site is designed around clear answers, useful resources and accessible content. Our own agency measurement must use questions asked by business owners choosing a provider—not homeowner questions from roofing experiments.
It is also the standard we suggest applying to anyone you hire, including us. Ask any visibility provider four things: how a citation is counted, what the denominator is, what happens to failed measurements, and whether you will see the full raw answers.
We have not published a measured improvement for Found & Proven. A technically sound site is a starting condition, not a success result.
Why these website foundations?
Google’s guidance emphasizes helpful original content and a clear technical structure for search and AI experiences. We use readable HTML, descriptive titles, internal links and accurate structured data. No special markup guarantees a citation.