← Back to Blog

When Your Entity Graph Splits Three Ways Because of a Naming Drift

When Your Entity Graph Splits Three Ways Because of a Naming Drift

LLMs disambiguate "which Acme are you" by cross-referencing the Organization / Person / Product entities you declare across your pages. When those declarations are consistent, the entity graph converges to one clean node and queries like "who is Acme" return your canonical definition. When they drift — even subtly — the graph splits into multiple weakly-linked nodes, and LLM retrieval scatters across them unpredictably.

The Entity Consistency Checker crawls up to 20 pages of a site, extracts every JSON-LD Organization / Person / Product object, every Open Graph tag, every sameAs profile, and cross-compares. Drift gets flagged. A unified-identity view tells you what the consistent version should look like.

The three flavors of drift

Name drift. The most common. "Acme Corp" vs "Acme" vs "Acme Corporation" vs "Acme, Inc." Often caused by different people writing different pages over time. Subtle enough that the site looks fine to a human reader but reads as three distinct entities to a parser.

URL drift. Org-level url property pointing to https://acme.com on one page, https://www.acme.com on another, https://acme.com/ (trailing slash) on a third. All three are different URLs to a strict resolver. LLMs flatten this most of the time; knowledge graphs don't.

Description drift. Three different one-liner descriptions for the same company. Each individually is fine. Together they give the LLM conflicting context about what the company actually does. If you want the LLM to describe you as "a storage company that builds portable units" and one page says "a logistics provider for SMBs," you've introduced uncertainty.

What the tool extracts per page

From each of the sampled pages, the checker pulls:

  • Every JSON-LD Organization, LocalBusiness, Corporation, Brand — with name, url, description, sameAs
  • Every JSON-LD Person — with name, sameAs
  • Every JSON-LD Product — with name
  • og:site_name from meta tags

Then it aggregates by value. If "Acme Corp" appears on 10 pages and "Acme" on 2, both show up in the output with counts. When a dict has more than one unique value, it's flagged as a conflict and the row borders turn red.

The sameAs dimension

Subtle but important: sameAs is the Wikipedia / Wikidata / social-profile link list on your Organization schema. It's how LLMs cross-reference your entity against external ground-truth sources. If your homepage lists 6 sameAs URLs and your About page lists 3 different ones, the LLM sees two partial views of your external footprint.

The tool aggregates sameAs across all pages and shows unique values with occurrence counts. The target: every page with Organization schema lists the same complete sameAs block. Use the E-E-A-T Generator to emit one canonical block and deploy it site-wide.

Why 20 pages

The tool samples the first 20 URLs from your sitemap. Most entity drift surfaces in 20 pages because it's the product of templating, not individual authorship — if the homepage template has one Org block and the blog template has a different one, sampling 20 catches both. For very large sites with many templates, run the tool against a different sitemap slice to catch template-specific drift.

The fix

For every conflicting row in the output, there's one canonical value and N deviations. Pick the canonical (usually the most-frequent; occasionally the most-correct). Update templates to emit the canonical. Re-deploy. Re-run the tool and the dict should collapse to one value per row.

Entity drift isn't fixed in one pass. Sites accumulate drift over years; a single pass gets the big wins. Plan a second pass three months later — new content tends to reintroduce drift if the template pipeline doesn't enforce the canonical.

Related reading

  • E-E-A-T Generator — emit canonical Person / Organization / sameAs JSON-LD to stop drift at the source.
  • Schema Completeness — complete JSON-LD graph per page type with linked @id.
  • NAP Consistency — explicit canonical-vs-listings NAP diff for local businesses.
  • Framework Origination — claim a coined method in the entity graph with DefinedTerm + CreativeWork.

Fact-check notes and sources


The $97 Launch covers how a coherent entity graph is the E-E-A-T foundation for an author site. The checker is how you verify every page carries the same coherent declaration.

← Back to Blog

Accessibility Options

Text Size
High Contrast
Reduce Motion
Reading Guide
Link Highlighting
Accessibility Statement

J.A. Watte is committed to ensuring digital accessibility for people with disabilities. This site conforms to WCAG 2.1 and 2.2 Level AA guidelines.

Measures Taken

  • Semantic HTML with proper heading hierarchy
  • ARIA labels and roles for interactive components
  • Color contrast ratios meeting WCAG AA (4.5:1)
  • Full keyboard navigation support
  • Skip navigation link
  • Visible focus indicators (3:1 contrast)
  • 44px minimum touch/click targets
  • Dark/light theme with system preference detection
  • Responsive design for all devices
  • Reduced motion support (CSS + toggle)
  • Text size customization (14px–20px)
  • Print stylesheet

Feedback

Contact: jwatte.com/contact

Full Accessibility StatementPrivacy Policy

Last updated: April 2026