How Ready Scan scores AI answers.

A scan starts from one public product page or homepage. Ready Scan proposes a brand, product, customer need and a pool of buyer questions. You can correct that interpretation and select one to four questions before confirming the scan. The questions define the market being sampled; they are not measured search-volume data.

For a discovery test, use questions a buyer could ask before knowing your brand. Keep the actual constraints, such as location, existing tools or a required capability. Removing a constraint just to make your brand appear changes the question rather than improving your position.

1. Check the target

   Confirm the correct brand, product and website. A company name, a product name and a target-owned domain can provide different evidence; a mention of one does not prove that an answer understood every offering.

2. Keep the sample comparable

   Record the selected wording, region and AI systems. A new question set measures a different set of opportunities, so a higher result is not automatically an improvement over the previous scan.

3. Read the observation date

   The date an AI answer was obtained matters more than when a report was opened. Where retained provider evidence is reused, its original observation time remains relevant; a new report date does not make the answer new.

Ready Scan retains the answer text and returned sources, then classifies the answer in the context of the buyer question. Classification is an interpretation that can be wrong; the answer and evidence let you inspect it. The arithmetic below is deterministic once those classifications are supplied. A narrative answer is the substantive text returned for a question, rather than a supporting list of search results.

1. Mention

   The retained answer names the supplied brand or product, or contains the target hostname. Name matching allows supported spacing and case differences, not invented aliases. A negative answer can still mention the brand.

2. Target citation

   The retained source list contains the target hostname or one of its subdomains, with ordinary www normalization. That is not necessarily a citation to the exact page you submitted. Open the cited URL before claiming a particular page was used.

3. Recommendation

   The answer makes the target a first choice, a clear positive recommendation or a conditional recommendation. Neutral inclusion in a list and a source citation do not, by themselves, count as recommendations.

4. Absence

   A completed, classifiable answer contains neither a target mention nor a target citation. An unanswered request, a failed request or an inconclusive answer is not brand absence.

Each eligible answer receives the points below. These are Ready Scan’s evaluation rules, not probabilities or an industry-standard measure. The two absence values distinguish an answer with a recorded direct competitor from one without one.

A score of 45 for absence is not a visibility win. It can exceed the 35 points for an incidental citation, which is why the point average must be read alongside the separate mention, citation and recommendation rates.

The AI Visibility Score is total outcome points divided by 100 times the number of scored answers, then multiplied by 100. Equivalently, it is the mean of their point values. Every scored question–AI pair has equal weight in this average. Results are rounded to one decimal place.

The denominator is eligible, completed, assessed narrative answers, not website pages, crawler visits, source links or every planned request. Supporting search-result observations and separately marked branded controls do not enter this average. With no eligible scored answers, the score and rates are unavailable, not zero.

Current reports lead with the Market Position Score. It combines the four components below, each expressed on a 0–100 scale: 0.50 × leadership + 0.25 × breadth + 0.15 × first-choice question coverage + 0.10 × AI Visibility Score. The final result is rounded to one decimal place.

This headline is not the percentage of answers that recommend you. Its weights express Ready Scan’s interpretation of the observed competitive position. If a retained report instead labels its headline AI Visibility Score, use the point-average formula above; the two numbers are not interchangeable.

A direct-competitor signal counts when that competitor is recorded in an answer where the target is absent, named but not recommended, or negatively evaluated. The same competitor must appear across at least two distinct buyer questions to qualify as recurring; several answers to just one question do not qualify. Adjacent workarounds are not direct competitors.

If several competitors qualify, the implementation selects the one covering the most distinct questions, then the most qualifying answers, then alphabetical label order to break a remaining tie. Let R be the target’s recommendation count and P that selected competitor’s qualifying-answer count. P is zero when no competitor qualifies. Leadership is calculated as follows.

1. R is positive and P is zero

   Leadership is 100. This means there is recommendation evidence without a recurring direct-competitor signal in this sample; it does not establish universal market leadership.

2. R and P are both positive

   Leadership is the smaller of 100 and (R ÷ P × 100). It is not the target’s share of all brands mentioned.

3. R and P are both zero

   Leadership is 50. The neutral starting value describes an unclaimed sample, not a recommendation or evidence that the brand was visible.

4. R is zero and P is positive

   Leadership is 0 because the sample contains a recurring direct-competitor signal but no recommendation for the target.

This is an illustrative calculation with fictional answer classifications, not a measured result from these providers. Suppose the two selected questions concern after-hours customer support and human handoff, and the scan obtains six narrative answers. Competitor R is a fictional direct competitor. Only the first-choice answer cites a target-owned source; the target is mentioned in the first-choice, conditional and neutral-list answers.

The point total is 100 + 75 + 10 + 10 + 60 + 45 = 300. With six scored answers, the underlying AI Visibility Score is 300 ÷ 600 × 100 = 50.0.

There are two target recommendations and two qualifying Competitor R answers across two different questions. Leadership is therefore 100. The target is visible in two of the three scored AI systems, so breadth is 2 ÷ 3 × 100. It is first choice for one of two questions, so first-choice coverage is 50.

The headline calculation is 0.50 × 100 + 0.25 × (2 ÷ 3 × 100) + 0.15 × 50 + 0.10 × 50 = 79.2 after rounding. Keep full precision for the breadth fraction until the final sum. A Market Position Score of 79.2 coexists with a recommendation rate of 33.3%; it does not mean 79.2% of answers recommend the brand.

Now change only the last Gemini answer from absent with 45 points to a completed but inconclusive answer. Six answers were returned, but only five can be scored. The 45 points leave both numerator and denominator: 255 ÷ 500 × 100 = 51.0, not 255 ÷ 600 × 100 = 42.5.

The mention, recommendation and target-citation rates become 60.0%, 40.0% and 20.0%. Leadership, breadth and first-choice coverage are unchanged, giving a headline score of 79.3. The apparent increases come from a smaller scored set, not an extra mention, citation or recommendation. Report both coverage and outcomes.

A missing or failed request is a different situation: current report assembly requires all selected answer and supporting-search requests to complete and every narrative answer to have an assessment before issuing the final report. A missing assessment also prevents completion. The scoring contract represents missing, failed and inconclusive observations separately; none is silently converted into brand absence. If every returned narrative answer is inconclusive, the score remains unavailable. First-choice question coverage still uses all selected questions, while an AI system with no scored answers is excluded from breadth.

The letter grade is a display band for the headline score. Apply the highest threshold reached in the table. These are product thresholds, not a percentile, an independent certification or a probability of a future recommendation.

The position label follows a separate rule: no target recommendations and no recurring competitor signal means unclaimed. Otherwise, a score of at least 80 with at least as many target recommendations as the leading competitor’s qualifying answers is leading; a remaining score of at least 65 is competitive; the rest are emerging. No scorable answers means unavailable. In the first worked example, 79.2 is grade A- and the position is competitive.

Use a score to decide which retained answers deserve attention, then inspect their wording and actual source URLs. The result identifies an observed gap; it does not reveal an AI system’s hidden ranking process or prove why it chose a competitor.

Keep the target, selected questions, region and AI systems consistent when comparing scans. Inspect original answer dates, source URLs, coverage and any disclosed evidence reuse. An automated observation of a selected answer service does not prove that every consumer-app session, retrieval setting or future answer will behave the same way.

The weights summarize a small selected sample, not the whole market. They cannot establish search demand, crawler satisfaction, indexing, website quality, referral traffic or sales. Even a high score does not guarantee a recommendation, and a score change alone does not prove that an edit caused it. Compare the retained answers and the concrete buyer needs as well as the numbers.

Ready Scan owns question planning, classification, score arithmetic and report presentation; answer providers supply the observed material. This reference documents the implemented market.position.v1 and visibility.score.v2 rules checked on 12 October 2026. The worked examples are reproducible arithmetic examples, not customer results or live provider tests.