Reading time
19 min
Date published
2025-07-15

SEO vol. 2 – How to Optimize Your Website in the Age of AI-Powered Search

The days when it was enough to land on Google’s first page are over. Search is evolving fast. More and more users are satisfied with AI-generated answers (like Google AI Overviews, Perplexity, or ChatGPT) — and they may never even visit your website.

Great content is no longer enough. Today, you need to create content that AI can both find and use in its responses. The SEO community initially reacted to AI search with fear. But over time, it’s becoming clear that these changes open the door to innovation — and a kind of “back to basics” approach that favors experimentation. That means a unique opportunity to get ahead of the competition.

In this article, we’ll break down the key principles behind this new mindset. And since a major German insurance company recently approached us for an AI-focused SEO strategy, we’ll also share some real-life examples from practice.

💡 AI search optimization is a must for any business that wants to stay visible. Those who start now will gain a massive competitive edge.

Let’s start with a stat. According to research, the average click-through rate (CTR) for informational queries dropped by 34.5% when AI Overviews were shown. Similarweb reported that from the launch of Google AI Overviews in May 2024 to May 2025, the share of so-called zero-click searches increased by 13 percentage points — from 56% to 69%.

What is AI search?

💡 AI search is an advanced way of finding information that uses machine learning and natural language models to understand the context of a query. It delivers more relevant, personalized, and comprehensive answers than traditional search engines.

How AI Finds Answers – and Why Traditional SEO Is No Longer Enough

When you enter a query into Google today — especially with the SGE (Search Generative Experience), ChatGPT, or Perplexity — you’re no longer just getting traditional search results. You’re shown an AI-generated answer: a summary of relevant information built by an AI model based on its understanding of the topic — not on exact keyword matches. And that completely changes the rules of the game.

💡 In the world of AI search, we’re not just optimizing for keywords anymore. We’re designing content that can be used across hundreds of possible queries. This approach is called Relevance Engineering.

1. Traditional SEO vs. AI Search

In the past, search engines focused on keywords. Algorithms matched search queries with website texts and ranked the best matches. Today, thanks to large language models (LLMs), it’s all about meaning. AI interprets a query as a user’s intent — not as a literal string of words. It looks for context, not exact matches.

Take these three queries:

  • “Best coffee machine for an office”

  • “Compare DeLonghi vs. Tchibo”

  • “Affordable coffee machine for company under €200”

AI interprets all of them as different ways of expressing the same need: to find a suitable coffee machine for the workplace. If your content meaningfully addresses that need — even if it doesn’t include any of those phrases word-for-word — AI can still use it to generate an answer.

2. Embeddings: How AI “Understands” Text

At the core of this understanding are embeddings — vector-based (numerical) representations of meaning. Instead of literal words like “office” or “coffee machine,” AI uses these vectors to measure how semantically close two texts are.

That means AI can recognize that:

  • “Affordable espresso maker” is close in meaning to “coffee machine under €200”

  • “Business supplies” relate to “office equipment”

  • “Brand comparison” is part of a purchase decision

So AI selects snippets not based on keyword repetition, but on context and relevance.

3. Query Fan-Out: How AI Searches for Content

Some systems — like Google SGE — use a method called query fan-out. That means the AI generates dozens of alternative versions of the original question to better capture what the user might be looking for.

For example, from the query “cheap coffee machine for the office,” AI might generate:

  • “Office espresso machine under €200”

  • “What’s best for coffee in an open office?”

  • “Capsule vs. manual machines — which one for business use?”

4. Selecting Relevant Content

Next, the AI scans available content (websites, articles, reviews, product descriptions) and picks out the most relevant snippets to build its answer. The clearer and more precise your content is, the more likely it is to be selected as a source.

5. How AI Understands Concepts: The Knowledge Graph

Google and some AI systems also use a knowledge grapha network of connected concepts, brands, locations, people, and topics.

For every word or phrase, AI builds context:

  • “Apple” might refer to a tech company or a fruit

  • “Jaguar” could be an animal or a car brand

To make sense of this, AI relies on the knowledge graph. For example, it “knows” that Apple is a brand that sells iPhones, is headquartered in the U.S., and has a CEO named Tim Cook — so it understands which Apple you mean.

If your website uses structured data (more on that below), clear terminology, and well-defined context, you’re helping AI place your brand correctly into this graph.

In our opportunity analysis, we focused primarily on visibility in AI-powered search — but traditional SEO is still essential. Comparing the client’s performance to competitors using classic organic search metrics helped us evaluate the overall quality of their website content.

Visibility in the traditional SERP still plays a significant role — at least in this early stage of AI search development — and often correlates with how well a website performs in AI-generated answers.
Real-Life Example
Source: Ahrefs
About us

How we can help you

Our team will help you with the first steps of your expansion as well as long-term marketing strategies.