Entity Analyzer
Audit your content for knowledge-graph anchored entities. We extract every named entity, resolve each against Wikidata, then flag the ghost references AI models cannot retrieve.
Audit Knowledge Graph Coverage
AI retrieval systems anchor on entities that exist in a public knowledge graph. An entity that resolves to a Wikidata Q-ID is a signal AI models can trust and cite. An entity that floats without an anchor is a ghost reference, present in your text but invisible to retrieval. This tool extracts every named entity from your content, resolves each one against Wikidata, then scores your Knowledge Graph Coverage. Ghost entities surface as action items with concrete fixes.
How It Works
- Paste a draft or published article into the text box.
- Click Run Coverage Analysis.
- We extract every named entity, query Wikidata for each, and score coverage.
- Copy the generated sameAs JSON-LD into your page schema.
Why It Matters
- Anchored entities resolve in Google's Knowledge Graph and most LLMs.
- Ghost entities float without referent and rarely earn citations.
- Strong coverage signals topical authority to AI retrieval systems.
- Branded or invented terms need anchors or expanded context to count.
What The Score Means
- 90 to 100: knowledge-graph dominant content
- 70 to 89: anchored, most entities resolvable
- 50 to 69: mixed, ghost entities present
- Under 50: floating, text without resolvable referents
Frequently Asked Questions
What is a "knowledge-graph anchored" entity?
An entity that exists as a record in a public knowledge graph, typically Wikidata. When AI retrieval systems read your content, they try to resolve every named entity to a known identifier. Entities with a Wikidata Q-ID resolve cleanly. Entities without one are unresolved tokens the model cannot fully trust or cite.
Why does Wikidata coverage matter for AI search?
Wikidata is the entity backbone of Google's Knowledge Graph and a primary training signal for most large language models. When your content mentions entities that exist in Wikidata, AI models recognize them, link them to surrounding facts, and weight your content higher in retrieval. Mentions of entities without anchors are treated as ambiguous strings.
My brand has no Wikidata entry. Is that a problem?
It is a real signal gap. Branded terms surface as ghost entities until you build an anchor. Options include creating a Wikidata entry directly, publishing a Wikipedia article (which Wikidata mirrors), or using
sameAs links to your own authoritative pages (About, Press, LinkedIn, Crunchbase) so AI models have any structured referent at all.What is a "ghost entity"?
A named entity present in your text that Wikidata cannot resolve. Common cases are branded product names, internal jargon, acronyms without expanded forms, and concepts new enough that no public knowledge graph has indexed them yet. The tool lists every ghost with specific fixes.
How does the tool handle ambiguous entities like "Apple"?
Wikidata returns multiple candidates for ambiguous strings. The tool flags these as ambiguous and surfaces the top three candidate Q-IDs so you can pick the right one when copying the
sameAs JSON-LD block. Production NER with context disambiguation lands in a later release.Is the analysis sent to a server?
Your text never leaves your browser. The tool runs entity extraction client-side, then queries the public Wikidata API for each unique entity. Wikidata sees individual entity names, never your full content.
MAY THE FORCE BE WITH YOU