EAC Score: The GEO Metric Predicting AI Search Citations

Executive summary (no rung‑1)

Citation is an identity problem before it’s a content problem. AI answer engines increasingly resolve entities (who/what is this) and only then attach claims. If your entity is ambiguous, you can publish great content and remain uncitable.
Entity Anchor Coverage (EAC) Score is a 0–100 audit for “citable identity.” It measures whether your brand is easy to resolve, verify, and attribute across independent surfaces.
EAC is not “add more schema.” It’s redundant identity wiring across: (1) your site’s entity dossier, (2) cross‑web identity nodes, (3) public entity graphs, and (4) machine consumption paths (sitemap/robots/llms.txt + canonical hygiene).

If you’re serious about GEO, you need a metric that penalizes identity uncertainty. EAC is that metric.

The contrarian premise: GEO is turning into entity resolution

Most GEO advice is document craft:

front‑load answers
add schema
write more FAQs
clean your HTML

Necessary, yes. But incomplete.

Google defines structured data as a way to give “explicit clues about the meaning of a page” and notes it uses structured data to understand the web and the world (including “people… or companies… included in the markup”). That’s an entity framing, not a “snippet markup” framing. (Google Search Central)

When engines aren’t confident they understand who is speaking, they become conservative about citing.

So the real question isn’t “Is my content good?”

It’s: “Is my entity easy to resolve, verify, and attribute?”

The EAC Score (0–100): what you’re scoring

EAC measures coverage + consistency of your entity anchors. Think of it as the opposite of “brand vibes.” It’s identity in machine terms.

Component A — On‑site entity dossier (0–25)

Do you publish a clean, machine‑readable identity card?

Minimum checks:

Organization (or Person) JSON‑LD exists on the canonical About/Company page (and/or sitewide)
stable @id (e.g., https://example.com/#org)
name, url, logo, contactPoint
sameAs links to authoritative profiles

Component B — Cross‑web “sameAs graph” (0–25)

Do independent sites corroborate the same identity?

authoritative social profiles (LinkedIn, YouTube, GitHub, X, etc)
relevant directories (G2/Capterra/Crunchbase where appropriate)
consistent brand name + URL + description
those nodes link back to the canonical site (where possible)

Component C — Public entity graph presence (0–25)

Are you present where graphs are built?

Wikidata entity exists (or is feasible) with references
Wikipedia only when legitimately notable (don’t spam)
a “KG mindset” check: would Google’s Knowledge Graph model find a stable entity for you?

(Useful mental model: Google’s Knowledge Graph Search API exists because entity lookup is a first‑class problem.)

Component D — Machine consumption paths (0–25)

Can machines ingest you cleanly?

/sitemap.xml exists, canonical, not noisy
/robots.txt doesn’t block important content by accident
/llms.txt exists, curated, and points to authoritative pages/markdown
canonical hygiene: 1 topic → 1 canonical URL (not 6 near‑duplicates)

Interpretation: EAC doesn’t reward “marketing effort.” It rewards removing identity uncertainty.

The audit method (60–90 minutes, repeatable)

1) Extract your on‑site entity objects

- Find application/ld+json blocks.

- Validate with Google’s Rich Results Test (it’s not a complete schema validator, but it’s a fast sanity check).

2) Build your identity node inventory

- official social profiles

- product listings / directories

- founder profiles (if person‑led brand)

3) Create a “sameAs matrix”

- rows = identity nodes

- columns = {name, url, logo, description, category}

- highlight inconsistencies (these are citation risk)

4) Score fixes by leverage

- anything that reduces ambiguity on high‑authority nodes outranks “write another blog post.”

Measured example (benchmark proxy): a quick EAC snapshot of geooptimizer.ai

I ran a lightweight check of the public site surfaces (homepage + machine endpoints):

https://geooptimizer.ai/ exposes 1 JSON‑LD block, type WebSite.
That JSON‑LD block contains no sameAs links.
https://geooptimizer.ai/llms.txt exists (HTTP 200) and is populated.
https://geooptimizer.ai/sitemap.xml exists (HTTP 200).
https://geooptimizer.ai/robots.txt exists (HTTP 200).

What this implies

This is already ahead of most sites because the machine paths exist (llms.txt + sitemap + robots).

But a WebSite object describes the site—not the organization behind the claims. If your “entity dossier” isn’t explicit, engines have to infer it from scattered hints. That’s where citations get conservative.

Example score (illustrative)

A) On‑site dossier: 10/25 (schema exists but missing explicit Organization + sameAs)
B) sameAs graph: 8/25 (likely exists, but not wired as a reconciled graph)
C) public entity graph: 10/25 (not assessed here; assume partial)
D) machine paths: 22/25 (strong: llms + sitemap + robots)

EAC ≈ 50/100.

Fastest path to 70+ isn’t “more content.” It’s identity wiring.

The 80/20 remediation plan

1) Publish an Organization JSON‑LD object (canonical About page; optionally sitewide)

Include:

stable @id URI (e.g., https://geooptimizer.ai/#org)
name, url, logo, contactPoint
sameAs list (only authoritative profiles)

Schema.org’s Organization type is the standard vocabulary for this. (Schema.org)

2) Make llms.txt point at the entity dossier

llms.txt is explicitly designed as a curated, LLM‑friendly overview—use it as an identity router:

link to About/Company page
link to “What we do” and “How we measure” pages
link to your strongest evidence assets (case studies, benchmarks, research notes)

3) Normalize your external nodes

Pick one canonical phrasing for:

brand name
one‑line description
category

Then make the same wording appear on the top identity nodes.

4) Create one “citation anchor page” per strategic theme

Not a blog post. An anchor page:

defines terms
states claims as atomic facts
cites sources
provides a stable URL worth citing

How GEOOptimizer Measures/Operationalizes This

GEOOptimizer operationalizes EAC as an authority system, not a checklist:

Schema extraction: detect whether Organization/Person objects exist and whether sameAs is populated.
Graph consistency: compare name/url/logo/description across identity nodes.
Machine paths: verify llms.txt, sitemap, robots, and canonical hygiene.
Remediation queue: produce a prioritized fix list (fastest ambiguity reduction first).

If you can’t measure “citable identity,” you can’t improve it. EAC makes the work legible.