Glossary
Entity optimization (for AI answers)
Same spirit as Knowledge Graph SEO, but tuned for generative retrieval and citations.
Pairs with schema.org, Wikidata-style references, and authoritative bios.
Definition
Entity optimisation focuses on the identity layer of your brand: official name variants, stock tickers, founder names, HQ locations, product lines, and how those facts appear in structured data and trusted third-party profiles. When entities are messy, models merge you with homonyms, invent subsidiaries, or swap pricing tiers. Cleaning entities raises the odds that RAG pipelines retrieve the correct chunk and that evaluators mark answers as wins.
Before/after disambiguation
| Signal | Before | After |
|---|---|---|
| Wrong competitor merges | 18% | 6% |
| Wrong HQ mentions | 12% | 3% |
| Identity prompt win rate | 52% | 78% |
Entity cleanup often gives faster gains than publishing new content.
How it's computed
There is no public formula — progress is observed indirectly via mention detection, duplicate entity counts in answers, and hallucination rates on identity prompts. Practitioners track consistency scores across fanout queries once canonical facts are published.
Consistency indicator
Identity consistency = correct entity facts / total entity fact checks
How it works in practice
Concrete tasks
- Publish
Organization/ProductJSON-LD with stable@idURLs. - Align Wikipedia/Wikidata, Crunchbase, App Store, and marketplace listings.
- Maintain a “do not confuse with” note for similarly named competitors.
Entity data card template
| Field | Example |
|---|---|
| Official name | Acme Cloud Inc. |
| Common short name | Acme |
| Product lines | Acme CRM, Acme Desk |
| Not to be confused with | Acme Logistics Ltd |
| Canonical source URL | /about/company-facts |
How to read it
If LLM-Score is volatile but content is strong, run an entity audit before rewriting articles.
Entity optimization sprint
- Build one canonical fact sheet page.
- Add organization and product schema with stable IDs.
- Harmonize brand facts across major profiles.
- Create 20 identity prompts for regression checks.
- Re-run monthly after any naming or portfolio change.
Identity quality bands
| Identity consistency | Reading |
|---|---|
| <65% | high confusion risk |
| 65-80% | partial clarity |
| 80%+ | strong entity hygiene |
Strong entity hygiene reduces hallucinations and improves recommendation confidence.
Do a regression check after every rebrand, merger, or new-market launch to prevent identity drift.
When to use
- After mergers, rebrands, or multi-brand portfolios.
- When models keep inventing offices you do not have.
- When launching in a new language market with duplicate transliterations.