What the score is
A 0–100 reading of how ready a business is to be described correctly by AI assistants — readiness, not results.
Measurement standard
The Answer-Readiness Score: HAVENGAI’s method for measuring AI visibility
This page documents the Answer-Readiness Score: HAVENGAI’s method for measuring how prepared a business is to be described correctly by AI assistants. It defines the seven inputs, the 0–100 output, and the prioritized fix list — and it states, explicitly, what the score does not promise. It is published as a standard because a measurement you cannot inspect is not a measurement.
What this gives you
The seven inputs
Seven observable, checkable inputs — each one defined below, none of them invented or estimated.
The output
The score plus a prioritized fix list: the actions ordered by what would move the reading most.
The boundaries
No ranking guarantees, no promised outcomes — platforms control their outputs; the score measures preparation.
What the score measures
The Answer-Readiness Score measures a business’s preparedness to be described correctly when an AI assistant answers a question about it — or about businesses like it. It does not measure rankings, traffic, revenue, or any outcome. It measures the state of the evidence: is the entity resolvable, corroborated, and current across the sources assistants read?
The distinction is deliberate. Rankings are positions on pages the platforms control; readiness is the condition of your own evidence, which you control. A business can improve its readiness completely and still wait on a platform’s crawl schedule. The score tells you how ready you are. It does not tell you when the platforms will notice.
The seven inputs
Each input is observable and checkable. Nothing in the score is estimated, projected, or modeled from proxies.
One — corroboration mass: the count of independent domains carrying consistent name, address, phone, and description for the business. Independent means not owned by the business; consistent means identical, not approximate.
Two — schema signal: the completeness of the business’s Organization structured data — name, address, phone, description, and identity links to owned profiles, present and valid in the page source.
Three — sitemap integrity: whether the homepage is listed in the published sitemap and whether the site’s pages are listed — the basic discoverability contract with crawlers.
Four — GBP state: the condition of the Google Business Profile — claimed, completed, consistent with the website, and verification status stated honestly rather than assumed.
Five — review footprint: the count and recency of reviews on third-party platforms — real customer reviews on independent surfaces, never fabricated or purchased.
Six — probe accuracy: what ChatGPT, Perplexity, and Gemini actually say when asked about the business, scored against the verified facts — correct industry, correct identity, no hedging disclaimers.
Seven — answer density: the percentage of the site’s pages that carry direct-answer blocks — plain, quotable statements of fact that assistants can lift accurately.
The output: a score and a fix list
The seven inputs produce a score from 0 to 100 and a prioritized fix list. The score is the reading; the fix list is the product. Fixes are ordered by what would move the reading most — the missing or weakest inputs first — so the business knows what to do on Monday morning, not just how it scored on Friday.
We do not publish internal weighting, score bands, or thresholds on this page, for a stated reason: the platforms’ behavior shifts, and a fixed formula would be precise about the wrong thing. The inputs are public and checkable; the fix list is actionable; the number is a summary, not a verdict. Any vendor selling you their exact formula as a guarantee should be asked what happens when the platform changes.
How it is used
The Answer-Readiness Score is computed inside the Position Scan — HAVENGAI’s AI-assisted review of a business’s market position, visibility, and lead capture. The scan gathers the seven inputs from observed evidence, produces the score and the fix list, and a human researcher reviews the findings in parallel with the automated checks before anything is delivered.
The score is a diagnostic instrument, not a product tier. It does not unlock features, gate pricing, or expire. It is re-computed when the evidence changes — after fixes are made — because a measurement of current readiness should reflect the current evidence.
What this does not promise
The score promises nothing about rankings, traffic, revenue, or platform behavior. Platforms control their outputs: their crawlers, their entity graphs, their answer rendering. No measurement of your readiness can compel a platform to act on a schedule, and we do not claim otherwise.
It does not promise that a high score produces correct answers immediately — crawl and reprocessing take the time they take. It does not promise that any specific fix produces any specific point gain; the fix list is prioritized by expected movement, not guaranteed movement. And it is not comparable across businesses as a competitive ranking: two businesses with the same score can face entirely different name competition and platform histories.
This section exists because the standard requires it. A methodology page that lists only what a method does is marketing. A methodology page that lists what it cannot do is a standard.
Evidence-bound by construction
Every input to the score is a fact that can be checked: a domain count that can be re-counted, schema that can be viewed in page source, a sitemap that can be fetched, a profile whose fields can be read, reviews with dates on them, assistant answers that can be re-asked, pages that either contain direct-answer blocks or do not.
That is the Jesus test applied to measurement: only what is observed goes in. If an input cannot be observed, it is not an input — it is cut. The score is therefore auditable by anyone willing to do the same checks, which is the entire point of publishing the method.
Your next move
Get your business’s Answer-Readiness reading.
Direct answers
What serious buyers ask.
Is a high Answer-Readiness Score a guarantee of correct AI answers?
No. The score measures the readiness of your evidence, not the behavior of platforms. Correct answers follow from strong evidence on the platforms’ own schedules — the score tells you how prepared you are, not when they will respond.
How often should the score be re-computed?
When the evidence changes — after identity fixes, new profiles, new reviews, or new pages are published. Re-scoring unchanged evidence on a schedule produces motion without meaning.
Which assistants are probed?
ChatGPT, Perplexity, and Gemini: each asked what the business is, with answers scored against the verified facts. One surface is anecdote; three surfaces are a reading.
Can I compute this myself?
The seven inputs are defined above precisely so you can. Count the corroborating domains, read your schema, fetch your sitemap, check your profile, count dated reviews, ask the three assistants, and audit your pages for direct-answer blocks.
Why don’t you publish the weighting formula?
Because platform behavior shifts, and a fixed public formula would be precisely wrong the moment it does. The inputs are public and checkable; the fix list is actionable. The number summarizes; it does not adjudicate.