AI Search Runs on Intent, Not Keywords

VZ Lens

The strategic shift is simple: search optimization is no longer about phrase matching first, but intent disambiguation first. The team that models intent structure better compounds durable AI visibility.

TL;DR

Keyword research (search volume, competition, CPC) is losing its leading role in strategy. AI overviews and chatbot retrieval respond to intent and context, not keyword density. Teams that still optimize mainly for monthly volume are optimizing a proxy that no longer tracks decision impact. The shift is toward intent clusters, topical authority, and quotable answer blocks.

---

I did my first keyword research eight years ago. I opened the Keyword Planner, entered the industry, looked at the monthly search numbers, saved them to an Excel file, and built a content calendar from that. The logic was simple and sound: the more people search for something, the more traffic a good ranking brings.

This logic still holds true today. However, the concept of “good rankings” has changed.

Today, there aren’t ten blue links at the top of the search results. There’s a summary—the AI Overview—which is the answer itself, not the path to the answer. Fewer people are clicking on the organic results below it than before. Chatbots—ChatGPT, Perplexity, Gemini—are increasingly replacing direct searches. People are increasingly asking questions in sentences, not keywords.

The Excel file I opened in 2017 measures something different today than what I want to measure.

---

Why doesn’t AI think in terms of keywords?

For decades, Google’s search algorithms were built around keyword matching—simply put, the page that more accurately contained the entered word or phrase ranked higher. This was the basis for keyword density, title tag optimization, and anchor text.

Large language models (LLMs) do something different. They don’t search for words—they search for semantic similarity. BERT, T5, and today’s transformer-based retrieval systems don’t ask, “Does this word appear in the document?” Instead, they ask, “Does this document address the same concept the user is seeking an answer for?”

This difference isn’t a technical detail. It’s a structural shift.

If someone types into Google: “how not to be tired after meetings”—classic keyword search would say: optimize for the keyword “meeting fatigue,” or perhaps variations like “energy management.” AI-based retrieval asks: what is the intent behind it? Is this an energy management question? A sleep question? A cognitive load question? Or perhaps an introversion question?

Content that precisely understands and addresses this intent will be cited. Content that merely contains the keyword will not necessarily be.

---

What is an intent cluster?

An intent cluster is a group of intents rather than a group of keywords. It is not a list that collects variations of a phrase—but a map that shows what questions, problems, and contexts lie behind a single underlying intent.

Let’s look at a concrete example.

Old approach — keyword list:

• “RAG architecture”
• “what is RAG”
• “retrieval-augmented generation”
• “RAG LLM”
• “using RAG”

These all refer to the same term. There is high overlap and low actual content diversity.

Intent cluster approach — covering the same topic, but organized by intent:

• Understanding intent: What is RAG? How does it differ from fine-tuning? When should you choose one over the other?
• Decision-making intent: Which vector database should I choose for RAG? What infrastructure is required?
• Implementation intent: How do I set up a RAG pipeline? What kinds of errors occur in a production environment?
• Evaluation intent: How do I measure RAG accuracy? What is the RAGAS metric?

The same topic—but four completely different pieces of content, for four different target audience segments. There is a different “ideal answer” for each intent. If you only write keyword variations, you leave most intents unserved—and the AI won’t cite you when searching for an answer to a question targeting that intent.

---

Topical Authority vs. Keyword Authority

In keyword-based SEO, the concept of “authority” was backlink-centric: how many external pages link to you, and what is their domain authority. This is still relevant today—but in GEO strategy, another dimension matters just as much: topical authority.

Topical authority measures how deeply and consistently a website as a whole addresses a given topic. AI retrieval systems—and Google’s BERT-based ranking—favor domains that cover a topic not superficially, but in a structured, multi-layered way.

This means that vargazoltan.ai isn’t cited in AI Overviews because it appears in a multitude of keywords—but because it covers the topics of AI strategy, GEO, enterprise RAG, and organizational AI implementation in a deep, coherent, and quotable format. All articles together build topical authority—they don’t compete individually for their own keywords.

Keyword competition is a competition between individual pages. Topical authority is a competition for an entire domain.

---

What changes in practice for content strategy?

Three specific shifts required by the GEO era:

1. Monthly search volume shifts from a primary metric to a secondary one. It doesn’t need to be abandoned entirely—but it can’t be the sole starting point. A query that “only” gets 50 monthly searches but is conveyed to thousands of people by AI chatbots in their summaries can be a more important target than a keyword with a monthly volume of 5,000 for which the AI only provides a generic answer.

2. Content intent compatibility is becoming more important than keyword coverage. Do you clearly articulate the intent that the visitor—and the AI—is looking for? If the article only has the right keywords but fails to fully address the intent, the AI won’t highlight it.

3. Contextual cross-linking (internal linking structure) is the visualization of the intent cluster. If your articles’ internal links follow the logic of the intent cluster—the page targeting the comprehension intent links to the decision-making intent article, and the implementation article links back to the architecture article—then the domain as a whole communicates topical authority to the AI.

---

What Hasn’t Changed

It’s important to note: the basic criteria for good content—accuracy, depth, readability, and source citations—have not lost their value. Quite the contrary.

AI retrieval systems prefer to cite texts that contain concrete data, make clear statements, and make sense on their own, even without context. This is exactly the kind of content a good technical writer would produce anyway.

The change isn’t that you now have to do something completely different. The change is that the old metrics—search volume, keyword rank, click-through rate—are becoming less and less effective at measuring what you’re actually striving for: relevance, citability, and brand recall.

Keywords are not dead. But keyword research as the sole navigation system is.

---

Key Takeaways

• AI Overview and chatbot searches respond to intent and context, not keywords—which is why optimizing keyword density yields increasingly lower returns
• An intent cluster is an intent map that maps all relevant questions, contexts, and target audience segments within a topic—this is the building block of GEO strategy
• Topical authority—a domain’s deep and consistent presence on a topic—is becoming an increasingly strong signal in AI retrieval selection
• Monthly search volume isn’t disappearing from metrics, but it’s becoming secondary to intent coverage and citability

---

Related thoughts

• How to Make a Website GEO-Friendly: 5 Concrete Steps — The basics of GEO and the most important technical steps
• Zero-Click Content Strategy — How to write in a way that gets people to quote you, not just click
• Entity-based SEO and GEO — Why AI thinks in terms of entities, not keywords

---

Zoltán Varga - LinkedIn Neural • Knowledge Systems Architect | Enterprise RAG architect PKM • AI Ecosystems | Neural Awareness • Consciousness & Leadership If AI doesn’t understand your intent, your keywords won’t save you.

Strategic Synthesis

• Map one core topic into 3-4 intent clusters before writing new content.
• Design self-contained answer blocks with explicit claims, evidence, and context.
• Measure authority through citation quality and decision influence, not traffic alone.