How GEOExtension scores a page

What does the page-readiness score measure?

Readiness is a deterministic audit of observable page signals. Every applicable check returns pass, warn, or fail. A check can also return N/A when the current surface cannot measure it or it does not apply.

pass = 1.0 · warn = 0.4 · fail = 0.0
readiness = Σ(weight × status value) / Σ(applicable weights) × 100

N/A checks are excluded from both sides of the calculation. The report exposes every result and category score behind the number. Readiness is a diagnostic—not a prediction of traffic, citations, revenue, or an AI “rank.”

Which readiness checks are included?

The current audit contains 21 checks. Higher weights reflect more direct barriers to retrieval and understanding; emerging conventions are labeled and intentionally carry little weight.

How is observed AI visibility measured?

A visibility probe asks a frozen, hashed set of representative buyer questions. Results are reported per provider as answered and failed questions; recommended, listed, mentioned, or absent; domain citations; observed entity position; and mention rate with a Wilson 95% confidence interval.

Live-web grounded answers and model-memory answers stay separate because they observe different systems. Runs with different question-set hashes are not presented as a trend. A provider timeout or error reduces the reported sample—it never counts as a brand absence.

How are recommendations prioritized?

Recommendations combine actual audit findings and provider observations using four explicit considerations: severity, evidence, confidence, and estimated implementation effort. Each recommendation should identify the affected page or observation, explain the consequence, and include a verifiable completion condition.

1. 1. Severity: does the issue prevent retrieval, understanding, citation, or recommendation?
2. 2. Evidence: is it supported by deterministic page evidence, provider observations, or both?
3. 3. Confidence: how directly does the evidence support the recommendation?
4. 4. Effort: what is the rough human implementation cost?

What are the limitations?

AI answers are sampled, non-deterministic observations—not stable rankings. Provider changes, location, personalization, retrieval availability, and time can change an answer. No audit score guarantees that a page will be indexed, cited, recommended, or produce business results. New heuristics begin as labeled experimental signals and should earn weight only after calibration against observed outcomes.