Free browser-local planning tool
Report a range, not a deceptively precise AI visibility score.
Enter your prompt panel, engines, repetitions and observed brand mentions. The calculator returns the observed mention rate, a 95% Wilson interval and a conservative sample-size plan. Nothing you enter is sent to Miando.
Plan the observation set
What the calculation does
Counts explicit outcomes
Each observation is one defined question, engine and run. A mention is recorded only when the answer meets your frozen rule.
Shows sampling uncertainty
The Wilson interval gives a range around the observed proportion without collapsing the result into a single precise-looking percentage.
Plans a conservative panel
The sample-size estimate uses the maximum-variance case, p = 0.5, at 95% confidence. It rounds up to complete runs across your prompt and engine panel.
n = z² × 0.25 ÷ margin² | z = 1.96 for a 95% planning level
Important limitations
- Generated answers from the same engine, prompt family or time window may be correlated. More rows do not automatically create independent evidence.
- The interval describes the observations entered under a binomial model. It does not prove future visibility, ranking, causation or business impact.
- Freeze the brand rule, prompt wording, engine, model, account state, market, language and collection window before comparing periods.
- For serious inference, review the sampling design with a qualified statistician and consider cluster-aware or bootstrap methods.
Primary references
- Quantifying Uncertainty in AI Visibility — repeated generative-search measurements, variability and confidence intervals.
- NIST Product and Process Comparisons — binomial proportion interval methods and worked examples.
Keep the raw observations beside the range.
Use the free workbook for evidence capture, or explore the complete Audit Kit when you need the full client workflow.