KVS SCORE · KAIS VISIBILITY SCORE
The KVS score: your place in AI answers,
measured under a written protocol.
A score out of 100, four axes with published weights, documented readings and the evidence behind them. This page sets out what the score counts, how it is collected and what it does not claim to measure.
Protocol written by Gontran Guermachi, founder of KAIS (KaiZen AI Strategy), Bangkok.
DIRECT ANSWER
What is the KVS score?
The KVS score (KAIS Visibility Score) is a score out of 100 that summarises a business's presence and position in AI engine answers, on a fixed panel of buying questions. It combines four weighted axes: Presence 30, Position 20, Top positions 30, Coverage 20. Every score rests on documented readings, two per question and per interface under standard services.
- A frozen panel of buying questions, each tied to a service, a client type, a language and a market.
- Named interfaces, set in the protocol before collection, per market.
- Two readings per question and per interface, in neutral sessions, with no fictitious majority.
- Four separate rates beneath the score: presence, recommendation, preference, accuracy.
- Evidence delivered: raw counts, competitors named, sources cited, collection conditions.
- No classic Google data in the score: rankings and traffic are reported separately.
01 / THE FORMULA
Four axes, published weights.
The KVS is a weighted average of four axes. The weights are fixed and identical for every client; only the question panel and the interfaces change from one client to the next.
KVS = 30 × Presence + 20 × Position + 30 × Top positions + 20 × Coverage
Each axis is expressed between 0 and 1; the resulting score runs from 0 to 100.
Presence
Share of readings in which the business is named.
Position
The business's rank among the providers cited, when it is named.
Top positions
Share of readings in which the business is named first.
Coverage
Share of panel questions on which the business is named at least once.
Top positions weigh as much as presence. Being the fifth name in a list is not worth the same as being the first name offered, and the weights reflect that.
Answer sentiment is recorded and reported, but it is not part of the score: a judgement of tone is less stable than a count. Nor does the KVS measure each of the nine dimensions of the KAIS AI Visibility Chain™ separately.
Two readings, always stated. As an absolute reading, the KVS is the weighted score above. As a relative reading, it is indexed to the panel leader: KVS = 100 × entity score ÷ leader score. The public barometers use the relative reading: a KVS of 100 marks the leader of the measured panel, not total visibility, and two relative KVS scores only compare within the same panel.
02 / BENEATH THE SCORE
Four rates, read separately.
A single score can hide two opposite problems. The four rates stay visible beneath the KVS because each calls for a different fix.
| Rate | What it counts | What a low rate signals |
|---|---|---|
| Presence | Readings in which the business is named, in any context | An access, corroboration or topic-coverage problem |
| Explicit recommendation | Readings in which the AI advises the business, beyond a mere mention | Evidence, ideal client or process not explicit enough |
| Preference in comparisons | Comparison questions in which the business is preferred over the competitors cited | Missing or unsourced differentiation |
| Accuracy | Descriptions consistent with the brand fact sheet | Inconsistent third-party profiles, a namesake, outdated information |
03 / COLLECTION
How each figure is produced.
The protocol is written and named before collection. A KVS figure only means something alongside its collection conditions.
- Panel. Three question sets: a frozen core set used to compare one wave with the next; an exploratory set; a brand monitoring set. The KVS tracked over time is calculated on the core set.
- Sessions. Neutral, logged out, without history.
- Readings. Two per question and per interface under standard services. If they disagree, the result is marked as variable, with no “two out of three” majority rule.
- Window. All interfaces are collected within one documented collection window, 72 hours where that protocol applies.
- Languages. In each priority market, the core set is collected in the local language and in English: strong visibility in English can hide an absence in Thai or French. For Chinese-speaking clients, the core set is also collected in Simplified Chinese.
- Models. Every major model update is logged. If it makes two waves incomparable, a new baseline is set rather than splicing two series together.
Default interfaces
| Market | Interfaces |
|---|---|
| Europe | ChatGPT, Google AI Overviews / AI Mode, Gemini |
| Thailand | ChatGPT, Gemini, Google AI Overviews, Meta AI |
| Chinese-speaking clients Custom-quoted module | Doubao (ByteDance), Qwen (Alibaba), DeepSeek, Yuanbao (Tencent), collected in Simplified Chinese |
Readings by service
| Format | Questions | Interfaces | Readings per question | Readings per wave | What the result supports |
|---|---|---|---|---|---|
| Free first look | 1 | 3 | 1 | 3 | An indicative signal, never a score |
| AI Search Diagnostic | 12 | 3 | 2 | 72 | A KVS baseline |
| GEO Foundation Sprint | 5 | 3 | 2, before and after | 30 per wave | The effect of fixes on the targeted questions |
| AI Search Advisory | 20 | 3 | 2, monthly | 120 per month | A comparable month-on-month change |
| Managed GEO Growth | 30 | 4 | 2, monthly | 240 per month | A comparable month-on-month change |
04 / THE EVIDENCE
The score is never delivered alone.
Every KVS score comes with the material needed to check it.
| Delivered | Detail |
|---|---|
| KVS score and its four axes | The score out of 100 and each axis's contribution |
| Four rates | Presence, explicit recommendation, preference in comparisons, accuracy |
| Raw counts | For example “named in 9 of 72 readings”, next to the percentage |
| Competitors named | Who the AI names instead of you or beside you, question by question |
| Sources cited | The pages and domains interfaces display as sources, when they display any |
| Collection conditions | Date, interface, language, market and session state for every reading |
| Recorded answers | The answer text, in full or as excerpts depending on the service |
| History | The series on the frozen core set, wave after wave |
Any figure can be re-run on the same question, the same interface and under the same conditions. The result may differ: that is the variability the two readings document.
05 / THE LIMITS
What the KVS score does not measure.
- Your Google rankings and traffic. The KVS contains no classic Google search data. Organic rankings, traffic and enquiries are reported separately, by confidence level.
- What every user sees. An AI answer varies with the account, location, history and model version. The KVS describes the conditions collected, not every user.
- The nine Chain dimensions. The KVS summarises presence and position; it does not replace the dimension-by-dimension reading of the KAIS AI Visibility Chain™.
- A guaranteed result. No provider controls a generative engine's answer. KAIS guarantees the protocol, not the score.
- A first look. The free first look (one question, one reading) gives an indicative signal and never produces a score.
06 / FREQUENTLY ASKED QUESTIONS
What clients ask about the KVS score.
What is the KVS score?
The KVS score (KAIS Visibility Score) is a score out of 100 that summarises a business's presence and position in AI engine answers, on a fixed panel of buying questions, based on documented readings.
How is the KVS score calculated?
As a weighted average of four axes: Presence 30, Position 20, Top positions 30, Coverage 20. The weights are fixed and identical for every client. In the public barometers, this score is indexed to the panel leader: a KVS of 100 marks the leader of the measured panel. Answer sentiment is reported separately and is not part of the score.
How many readings per question?
Two per question and per interface under standard services, in neutral, logged-out sessions. If they disagree, the result is marked as variable; no “two out of three” majority is applied. A custom engagement may follow another protocol.
Does the KVS score include my Google rankings?
No. The KVS is calculated only from the answers of AI interfaces. Organic rankings, traffic and enquiries are reported separately.
What do I receive with a KVS score?
The four axes, the four rates, raw counts, competitors named, sources cited, collection conditions and the recorded answers, in full or as excerpts depending on the service.
Does the KVS score cover Chinese AI engines?
Yes, through a custom-quoted module for businesses serving Chinese-speaking clients: Doubao (ByteDance), Qwen (Alibaba), DeepSeek and Yuanbao (Tencent), collected in Simplified Chinese. Access conditions (account, location) are named in the protocol, because these interfaces can answer differently outside mainland China.
Can KAIS guarantee a KVS score?
No. No provider controls the answers of a generative engine. KAIS guarantees the measurement protocol and its documentation, not the result.
→
Get your KVS baseline.
The diagnostic: 12 questions, 3 interfaces, 2 readings per question, and three priority hypotheses.
Not ready yet? The free first look: one question, three interfaces, no score.