Entities for Technical People: Knowledge Graphs, Tokens, and Query Fan-Out

In my last two posts, I explained entities at a high level.

Now let’s go deeper.

This one is for the engineers.
The developers.
The SEO professionals.
The technically curious business owners.

Because what’s happening underneath AI search is not magic.

It’s structure.

First: AI Is Not “Searching”

Traditional search engines match queries to indexed pages.

Large Language Models do something different.

They:

  1. Break a question into smaller components
  2. Convert it into tokens
  3. Predict likely next tokens
  4. Pull supporting information from indexed knowledge
  5. Generate an answer

That process relies heavily on entities.

Not keywords.

Entities.

What’s Actually Happening Under the Hood

When someone asks:

“Who’s the best commercial printer in New Hampshire that specializes in short-run booklets?”

AI doesn’t look for that exact sentence.

It decomposes the query into entity clusters:

  • Commercial printer
  • New Hampshire
  • Short-run booklets
  • Trust signals
  • Reputation
  • Industry specialization

This is often referred to as query fan-out.

The system expands the query into related sub-queries and entity relationships.

It then evaluates:

  • Which businesses are strongly associated with those entities?
  • Which ones have consistent data across sources?
  • Which ones appear in trusted knowledge graphs?
  • Which ones have strong review density?
  • Which ones demonstrate authority?

This is why generic content fails.

If your website loosely mentions printing but doesn’t strongly associate itself with “short-run booklets” as a distinct entity, you’re weak in that graph.

Knowledge Graphs Matter More Than You Think

AI systems rely on knowledge graphs.

A knowledge graph is essentially a map of entities and how they relate to one another.

Business → Offers → Service
Business → Located in → City
Business → Has credential → Certification
Business → Reviewed by → Customer

The stronger and more consistent those relationships are, the more confident the model becomes.

Confidence determines citation.

And citation determines visibility.

Why Structure Is Non-Negotiable

If your site:

  • Has multiple H1s
  • Jumps heading levels randomly
  • Lacks a defined main content area
  • Mashes services into one vague page
  • Has inconsistent NAP data
  • Lacks schema markup

You’re weakening entity clarity.

😎 Humans can compensate for messy structure.

👉 Machines do not.

This is why at Toto SEO we obsess over:

  • Entity mapping
  • Site mapping
  • Structured data
  • Schema labeling (Organization, Service, FAQ, Review, Product)
  • Reader-mode compatibility
  • Business Facts Pages
  • Facts Blocks
  • Consistency across platforms

We are not optimizing for keywords.

We are strengthening entity relationships.

Tokens, Predictive Modeling, and Why Keywords Alone Fail

LLMs operate on probability.

They predict likely next tokens based on training data and contextual associations.

If your business is strongly connected to:

  • Specific service terminology
  • Geographic specificity
  • Industry vocabulary
  • Credible sources
  • External mentions

You are statistically more likely to be surfaced in generated answers.

If you are vague, inconsistent, or overly promotional…

You are weak in the probability model.

That’s not philosophical.

That’s math.

The Real Technical Shift

Old SEO focused on:

Match the phrase → Rank the page → Get the click.

New AI visibility focuses on:

Define the entity → Strengthen its relationships → Increase model confidence → Be cited in the answer.

That is a structural shift.

And most agencies are still operating in the old framework.

Practical Technical Takeaways

If you are technical, here’s your checklist:

  • One H1 per page, logically structured H2/H3 hierarchy
  • Clearly separated service pages
  • Organization schema implemented properly
  • Service schema for distinct offerings
  • FAQ schema for direct answer extraction
  • Review schema with specific, contextual language
  • Consistent NAP across all external profiles
  • Fresh content updates (XML sitemap reflecting last modified dates)
  • Author bios with credential signals
  • Internal linking reinforcing entity relationships

This is not cosmetic SEO.

This is architectural.

Why This Matters

AI tools are generating answers.

Not lists of links.

If your entity structure is weak, inconsistent, or shallow, you will not be confidently included.

If your entity relationships are strong and clear, you increase the probability of citation.

And in a world moving toward zero-click answers…

Citation is everything.

Next post, we’re going to zoom back out and talk about something bigger:

Why this moment feels exactly like the early 2000s when businesses didn’t understand why they needed a website!!

History is repeating itself.

And the early movers will win again.

Jennifer DeRosa

Jennifer DeRosa

Jennifer DeRosa is an AI-forward SEO strategist and author of Building DIY Websites for Dummies (Wiley).

She is the founder of Toto SEO, a GEO/SEO agency helping small businesses stay visible in both AI-driven and traditional search, and Toto Coaching, which provides DIY guidance for building credible, conversion-ready websites.

With 20+ years of experience, Jennifer built and sold her web development agency, TechCare (2001–2021), and completed MIT’s No-Code AI & Machine Learning program.

She is a frequent SCORE speaker and mentor, translating shifts in AI search into actionable strategies like entity-based optimization and structured data so businesses can be cited and trusted in ChatGPT, Google, and beyond.

Before forming TechCare, she consulted for companies including Mercedes-Benz Credit, U.S. Surgical, GTE, GE Capital, Unilever, and Calvin Klein.

Her work is known for measurable results, transparency, and ethical, standards-based implementation.

Ready to Stay Visible in AI Search?

The SEO landscape has changed. Are you ready to change with it?

Let’s talk about how Toto SEO can help your business stay visible, trusted, and competitive in the age of AI.