The KSM Model™ Score — Methodology
The KSM Model™ Score is a 100-point composite built from three pillars, weighted 40 / 40 / 20:
| Pillar | Weight | Measures |
|---|---|---|
| E — Structured Extractability | 40 pts | Can AI understand you? |
| S — Entity Salience | 40 pts | Can AI recognize you? |
| C — Citation Authority | 20 pts | Can AI trust you? |
Extractability and Salience are weighted evenly and heavier than Citation Authority, because a source can't be trusted or cited until it's first machine-readable and correctly identified. Citation Authority carries less raw weight but is the hardest to earn — third-party validation can't be self-published.
Why 40/40/20, and why it isn't just addition
The displayed score is a sum of the three sub-scores (e.g. 22 + 14 + 5 = 41/100), but the underlying model is multiplicative, not additive: AI Visibility = f(E × S × C).
That distinction matters. In a multiplicative system, a near-zero pillar caps the entire composite regardless of how strong the other two are — you can't average your way to visibility with two strong pillars and one collapsed one. The 100-point sum is a readable proxy for people; the multiplicative logic is what actually explains why a 41 behaves more like a "weak" score than a simple average would suggest, and why fixing the lowest pillar is always the highest-leverage move.
Pillar 1 — Structured Extractability (40 pts)
Whether machines — search crawlers, answer engines, non-rendering AI bots (GPTBot, PerplexityBot, Applebot) — can parse and lift your content cleanly.
| Sub-criterion | Points | What's checked |
|---|---|---|
| Schema.org / JSON-LD validity | 10 | Organization, Person, FAQPage schema present, valid, and error-free; nested @id linking (e.g. founder → ORCID) |
| Crawlability & indexation | 8 | robots.txt validity, sitemap status, meta-robots directives, confirmed indexation in Search Console / Bing Webmaster |
| On-page technical hygiene | 12 | H1 present, title tag length (~60 char), meta description length (~155 char, non-duplicated across meta/OG/Twitter), lang tag, image alt text |
| AEO-ready content structure | 10 | Dense FAQ blocks, clear headers, Q&A-formatted content that matches its own schema markup |
Common score-killer: structured data that's real but injected via JavaScript (e.g. Google Tag Manager) — invisible to crawlers that don't render JS. Valid schema that non-rendering bots can't see effectively scores as if it doesn't exist.
Pillar 2 — Entity Salience (40 pts)
Whether the entity itself — a person, org, or brand — is consistently and unambiguously represented across the knowledge graphs AI systems actually draw from.
| Sub-criterion | Points | What's checked |
|---|---|---|
| Knowledge graph presence | 14 | Live Wikidata node, Wikipedia article — the two signals AI knowledge-graph retrieval weights most heavily |
| Identifier consistency | 10 | ORCID, DOI, Scholar ID, sameAs linking between on-site schema and external profiles |
| Cross-platform presence | 8 | Institutional bios, LinkedIn, ResearchGate, Academia.edu — third-party-hosted anchors, not just self-published ones |
| Disambiguation | 8 | Whether the name/brand is unique enough in the corpus, or collides with unrelated prior use of the same term/acronym |
Common score-killer: no Wikidata node. It's the single highest-leverage line item in this pillar — its absence alone can hold Entity Salience under 50% even with strong identifier coverage everywhere else.
Pillar 3 — Citation Authority (20 pts)
Whether independent, third-party sources actually select and reference you — the signal that can't be self-published into existence.
| Sub-criterion | Points | What's checked |
|---|---|---|
| Peer-reviewed / indexed citation | 10 | Scopus/Web of Science-indexed citation of the underlying work — not just distribution on a preprint network |
| Independent third-party application/citation | 6 | Unaffiliated parties applying, referencing, or building on the work — and critically, whether that reference is itself indexed/discoverable |
| Press / industry citation | 4 | Media, competing vendors, or industry press referencing the entity or framework |
Common score-killer: real third-party validation existing but not yet linked from the source's own site or indexed anywhere crawlable. Unlinked proof contributes zero signal — evidence has to be discoverable to count.
Worked example — Dr. Anthony Q. Bowen / KSM Model™, July 2026 assessment
| Pillar | Score | Read |
|---|---|---|
| Structured Extractability | 22 / 40 | Strong schema and FAQ structure, undercut by five unresolved technical defects (missing H1, oversized meta, missing lang tag, unalt'd images, JS-only schema injection) |
| Entity Salience | 14 / 40 | Solid identifier trail (ORCID, Scholar, Zenodo) but no live Wikidata node, no Wikipedia, and a name-collision risk with unrelated prior "KSM" usage |
| Citation Authority | 5 / 20 | One genuine independent application exists but isn't yet linked or indexed; no peer-reviewed citation yet |
| Composite | 41 / 100 | Stage 2 — Structured Presence |
The takeaway the multiplicative model is built to surface: Structured Extractability is the strongest pillar here, but it can't lift the composite out of "weak" territory on its own — the Entity Salience and Citation Authority gaps are what's actually suppressing AI visibility, and they're where the next 60 days of work should go.