Optimizing Content for AI: How to Go From ‘Strings to Things’ and Pivot to Entity-Based SEO

In the world of high-level SEO, we stopped caring about “keyword density” a decade ago. We know that stuffing words onto a page is a relic of the past.

However, many sophisticated marketers are still trapped in a “Lexical” mindset. We still build traditional keyword research, hunt for exact-match search volumes, and judge a page’s success by whether it ranks for a specific string of text.

Now, the search landscape has fundamentally broken away from that model. With the rise of Large Language Models (LLMs) and Google’s Search Generative Experience (SGE), the algorithm is no longer just matching text strings; it is analyzing intent, context, and relationships.

The new signal of dominance isn’t about having the right keywords; it’s about having Topical Authority. To win, we have to make the strategic shift from “Strings to Things.”

Part 1: The Evolution (Why Keywords Stopped Working)

To understand where we are going, we have to understand why the old model broke. For years, Google was essentially a fancy digital filing cabinet. It matched the words on the label (your query) to the words in the file (the webpage). If you had the most matching words, you won.

Over the last decade, Google has been slowly “waking up” and learning to read like a human.

The 4 Major Shifts

1. The Knowledge Graph (2012): This was the birth of “Entities,” which Google defines and structures through systems like its Knowledge Graph. Google stopped just indexing strings of text and started mapping real-world objects. It began to understand that “Barack Obama” wasn’t just two words, but a person, a former president, and a husband.
2. Hummingbird (2013): The “Meaning” Update. Before this, Google looked at words individually. Hummingbird allowed Google to analyze the entire query at once, understanding the context of the sentence rather than just the sum of its parts.
3. RankBrain (2015): The “Intent” Update. This introduced Machine Learning. Google started guessing the intent behind ambiguous searches. If you searched “apple,” RankBrain used context clues to decide if you meant the fruit or the iPhone.
4. BERT (2019): The “Nuance” Update. BERT (Bidirectional Encoder Representations from Transformers) allowed Google to understand the nuance of prepositions like “to,” “for,” and “without.” It was the bridge to true natural language understanding.

Today, we are in the era of SGE (Search Generative Experience). The search engine doesn’t just want to find the document; it wants to read and synthesize the document to generate an answer. If your site is just a bag of keywords without semantic structure, the AI can’t use you.

Part 2: The Core Shift—”Strings” vs. “Things”

This is the most critical concept for modern SEOs to grasp.

• Keywords are “Strings”: A sequence of characters (e.g., “b-a-n-k”).
• Entities are “Things”: Distinct concepts—people, places, ideas, or brands—that the AI recognizes as objects with defined relationships.

When a user searches for SEO Consultant Toronto, a lexical engine looks for those letters. A semantic engine (AI) looks for the Entity.

It sees “SEO Consultant” (Job Role) and “Toronto” (Location). It then looks for the relationship between them. It asks: Does this website have the authority and the semantic connections to prove it is the best answer for this specific entity relationship?

If you rely on hitting keyword targets but fail to map out the relationships between these entities, the AI deems your content “thin.” You might have the keywords, but you lack the context.

Part 3: Real-World Example (Old vs. New)

Let’s look at a practical example. Imagine we are optimizing a page for a Personal Injury Law Firm.

Feature	The Old Way (Lexical SEO)	The New Way (Entity SEO)
Headline	Best Car Accident Lawyer in Toronto	Understanding Your Rights After a Motor Vehicle Accident in Ontario
Strategy	Repeat “Car Accident Lawyer” 5 times.	Cover related entities like SABS, Tort Claims, and Insurance Deductibles.
Content	Shallow content focused on sales.	In-depth resource connecting the “Lawyer” entity to the “Rehabilitation” entity.
Result	Ranks only for exact match queries.	Ranks for thousands of long-tail queries and Voice Search questions.

Why the Old Way Fails:

The “Old Way” reads robotically and risks “keyword stuffing” penalties. More importantly, it fails to answer the next question the user has. AI models predict the next logical token; if your content doesn’t cover the logical next steps (e.g., “How much is the deductible?”), the AI views your content as incomplete.

Part 4: Technical Implementation (The Schema “Secret Weapon”)

You cannot talk about Entity SEO without talking about Structured Data. While writing better SEO content is step one, step two is speaking Google’s native language: JSON-LD Schema, based on standards defined by Schema.org.

To firmly establish your brand as an Entity, you need to explicitly tell Google who you are using the sameAs property. This property acts as a “digital passport,” confirming that your website is the same entity as your LinkedIn profile, your Wikipedia page, or your Crunchbase listing.

Example 1: Defining a Person Entity

Here is how we might define the “Paul Teitelman” entity in the code of the homepage.

Example 2: Defining an Organization

For a business, we use the Organization schema to link the website to its physical location and social profiles.

By adding this code, we are not leaving it up to Google to guess. We are explicitly stating: “This website, this LinkedIn profile, and this person are all the same Entity.” This disambiguation is critical for showing up in Knowledge Panels and AI answers.

Part 5: The Execution—How to Find Entities (Step-by-Step)

You don’t need expensive software to find the entities Google associates with your topic. The data is publicly available if you know where to look. Here is the 3-step workflow we use during our Technical SEO Audits.

Step 1: The Wikipedia “Concept” Method

Wikipedia is the primary training ground for Google’s Knowledge Graph.

1. Go to the Wikipedia page for your core topic (e.g., “Search Engine Optimization”).
2. Look at the Table of Contents. These headers are usually the sub-entities you need to cover.
3. Look at the Internal Links in the first paragraph. If Wikipedia links “SEO” to “Web Traffic,” “Algorithms,” and “Organic Search,” those are the semantic relationships you need to mirror in your content.

Step 2: Google Images “Bubble Tags”

Google Images often reveals entities better than text search.

1. Search for your keyword in Google Images.
2. Look at the row of “bubbles” (related search tags) at the top of the results.
3. These bubbles represent the visual entities Google strongly associates with your topic. If you are writing about “Kitchen Renovations” and the bubbles say “Modern,” “Farmhouse,” “Cabinets,” and “Backsplash,” your content is incomplete without those sections.

Step 3: Google Trends “Related Topics”

Most people use Google Trends for keywords, but the real gold is in “Topics.”

1. Type in your keyword.
2. Scroll down to the bottom right box labelled “Related Topics”.
3. Ignore “Related Queries.” Look at “Topics.” These are the broad entities that users who search for your term are also interested in. Covering these helps you capture the wider user intent.

Part 6: Why AI Rejects “Thin” Content (The Vector Space)

How does an AI know if your content is “good”? It uses Vector Space.

Imagine a massive 3D map of the universe, but instead of stars, the dots are words. Words with similar meanings hang out close together in clusters. “Dog” is close to “Puppy,” “Bone,” and “Leash.” It is far away from “Spatula” or “Tax Return.”

When an AI analyzes a query like “Commercial Mortgages,” it looks at the vector space around that term. It expects to see related dots like LTV ratios, amortization periods, industrial zoning, and bridge financing.

If those dots are missing from your article, the AI views your content as “empty” or “thin.” You might have the keyword “Commercial Mortgage” in your H1, but because you lack the surrounding vector context, the AI knows you aren’t a true authority.

This is why Topical Authority is the new ranking factor. It proves to the algorithm that you possess a deep, 360-degree understanding of the subject matter, aligning with how Google evaluates content quality through its Search Quality Evaluator Guidelines.

Part 7: The Strategic Pivot—3 Steps to Update Your Approach

Changing your strategy requires a shift in how we structure content. At PaulTeitelman.com, we are moving clients away from “keyword hunting” toward “Entity Optimization.”

1. Optimize for Information Gain

In a world flooded with AI-generated content, “me-too” articles that regurgitate the same 500 words as your competitors are toxic to your rankings. Google’s recent patents prioritize Information Gain:  content that adds something new to the corpus of knowledge.

The Pivot: Stop asking “What keywords do I need to include?” and start asking “What unique data, expert perspective, or proprietary statistics can I add that no one else has?”

2. Adopt a “Hub and Spoke” Architecture

You can’t establish Topical Authority with a single page. You need to create a cluster of content that covers the breadth and depth of the entity.

The Pivot: Instead of five disconnected blog posts, create one authoritative “Pillar Page” (The Hub) regarding a core service—such as SEO strategy—and link it to supporting articles (The Spokes) that answer specific, long-tail questions. This signals to the AI that you cover the whole topic, not just the head term.

3. Shift from “Headlines” to “Natural Language Queries”

Voice search and chatbots have changed how users query. They ask questions; they don’t just bark keywords.

The Pivot: Review your H2 and H3 headers. Are they generic labels like “Our Services”? Change them to reflect the actual questions your users are asking, such as “How does Entity SEO differ from traditional optimization?” This structure makes it easier for AI to extract your content as a direct answer (Answer Engine Optimization) — an approach that aligns with broader AEO and GEO strategies.

FAQ: Entity SEO & The Future of Search

Do keywords still matter in 2026?

Yes, keywords are still the “anchor” points for search, but they are no longer the entire boat. You use keywords to signal the topic, but you use Entities to signal authority. In modern SEO, keywords help introduce the topic, while entities and context help search engines understand depth, relationships, and intent behind the content.

How do I know if I have “Topical Authority”?

If you rank well for a broad “head term” (e.g., “SEO”) but fail to rank for the granular long-tail questions (e.g., “how to do keyword research for SaaS”), you likely lack topical authority. You need to fill those content gaps. Strong topical authority means your content consistently ranks across related subtopics, showing search engines that you cover the subject in a complete and meaningful way.

Can AI write my Entity content for me?

AI is great at finding entities, but it is bad at “Information Gain.” If you use AI to write the content, you are just regurgitating the average of what already exists. You need to inject your own expertise and data to rank. Without original insights, case studies, or unique perspectives, AI-generated content often lacks the depth required to build true authority in competitive search environments.

Conclusion: Future-Proofing Your Rankings

The shift from Lexical targeting to Entity-Based SEO isn’t just a trend; it’s the necessary evolution of search. Google is building a map of the world’s knowledge, and it favours the sites that help it connect the dots.

If you are still counting exact match phrases in your headers, you are optimizing for 2015. To win in 2026 and beyond, you need to stop chasing strings and start owning topics.

Need to pivot your strategy?

If your organic traffic has plateaued, you may be stuck in a lexical trap. Contact me today to discuss how we can build an Entity-First strategy for your brand.

📚 Further Reading & Resources

To understand the technical mechanics behind this shift, I recommend exploring the following resources:

• What is Google Hummingbird? (Moz): https://moz.com/learn/seo/google-hummingbird
  A definitive guide to the 2013 update that shifted search from “strings” to “things.”
• Google Patent US20200349147A1 (Contextual Estimation of Link Information Gain):
  https://patents.google.com/patent/US20200349147A1/en
  The technical patent reveals how Google values unique information over repetitive content.
• BERT: Pre-training of Deep Bidirectional Transformers (Google Research): https://arxiv.org/abs/1810.04805
  The original research paper behind BERT, detailing how Google advanced natural language understanding to better interpret context, nuance, and user intent in search queries.