GEO in Action - How AI Started Recommending Us

Your SEO is working. You have high-quality content, strong rankings, and stable traffic. Yet, AI engines (like ChatGPT or Google AI Overviews) still ignore you in their recommendations. The problem isn’t that your writing is bad. The problem is that the algorithm doesn’t understand who you are. It cannot accurately categorize your expertise and map it to your website.

The result? Your potential client conducts research using AI, asks for industry experts, and receives a list of companies that doesn't include you.

This hurts the most in the B2B segment. Here, investments are high, and decisions take a long time to mature. If a buyer is doing initial research via AI today, this is the exact moment they decide whether to even invite you to the RFP (Request for Proposal). If the AI fails to mention you, the potential client won’t even include you in their shortlist.

The data backs this up - according to data publicly shared by Vercel CEO Guillermo Rauch in April 2025, ChatGPT referrals already account for 10% of new registrations on Vercel, up from less than 1% just six months prior.

In this article, using our own website as an example, we will show you what we changed so that AI would start recommending us with confidence, and how you can do the same.

From "Keywords" to Authority Building

Google has long since ceased to function merely as a keyword search engine. Today, it acts more like an experienced advisor trying to understand the context:

Who are you? → What do you truly understand? → In what topic are you so proficient that it can recommend you with a clear conscience?

Behind the scenes, algorithms build an "expert map" (technically known as a Knowledge Graph). This is a digital network of relationships between companies, their experience, and their results. If you don't have a clear and solid place in this network, AI will bypass you when generating responses. This is not discrimination; the algorithm simply prioritizes a source about which it knows more context, giving it greater confidence.

This experience is also backed up by an Ahrefs analysis across 75,000 brands: the correlation between the number of branded mentions on the web and visibility in AI overviews reaches 0.664. Traditional backlinks score only 0.22. In other words: AI doesn't ask who has the best SEO. It asks who it knows enough about.

That's why today, it's no longer enough to just "produce" good content. The algorithm needs to see that this content comes from someone with clear and consistent expertise. It's not enough to write about a topic; you must appear to the AI as a verified authority that genuinely understands the subject.

Key Takeaway: Today, AI evaluates text based on relationships, not keywords. Your success is determined by the "expert map" - an invisible network where the algorithm connects your company with specific services and real experts. If the algorithm doesn't see these connections clearly, it lacks the confidence to recommend you.

AI Visibility is a Strategy, Not a Coincidence (GEO)

This approach is called GEO (Generative Engine Optimization). It’s not about shortcuts or tricks; it’s about giving the algorithm enough confidence to cite you.

That these are not empty claims is proven by research from scientists at Princeton, Georgia Tech, and IIT Delhi titled "Generative Engine Optimization." Their findings showed that websites applying GEO principles saw up to a 40% increase in visibility within AI responses compared to standard content. The principle is clear: it’s not enough to be good; you need to be easily citable.

AI Doesn't Read Your Website Like an Article – It Looks for Facts

Artificial intelligence doesn't look for flowery sentences; it looks for clear relationships:

This company provides this service → they have years of experience in it → they use this specific technology to deliver it.

If these connections are not clearly defined within your code, the AI has to guess. However, when it comes to high-ticket B2B services, AI doesn't take risks – it prefers competitors about whom it possesses more precise data. This is where the technical layer of your website comes into play: data hidden in the code that explicitly tells machines who you are.

According to an AirOps analysis covering thousands of AI citations, only 30% of brands remain consistently visible across multiple LLM responses. Just because an AI mentioned you today doesn't mean it will mention you tomorrow for the exact same query. Stability only comes with consistent signals: the technical layer of the website, the density of facts within the content, and mentions across external sources.

Key Takeaway: GEO is all about giving the algorithm enough certainty to cite you. This is achieved through the technical layer of the website (structured data) and a consistent presentation of expertise within the text. If the AI has to guess or fill in the blanks, it will choose to recommend a competitor with more precise data for B2B services.

Case Study: What It Looks Like in Practice

The following examples are taken directly from our own website. Please note that these are not guaranteed results. AI overviews are not static; they change constantly based on the user query, context, and algorithm updates. The screenshots capture a specific moment in time. However, the underlying principle behind them remains valid in the long run.

Example 1: Headless CMS

We explicitly defined our expertise in this technology directly within our source code.

The Result: Today, Google AI automatically ranks us among the top agencies in Slovakia. Although we already rank high in traditional SEO, the AI overview appears even higher on the page – and we are a direct part of it.

Example 2: Server-Side Tracking

Here, the difference is even more striking. Although we don't hold the absolute top spot in traditional organic search results, the AI evaluated us as a key player thanks to the clearly defined context in our code – and it recommends us above all other standard links.

Both examples demonstrate the exact same thing. Getting featured in an AI overview isn’t determined by your Google ranking. It is determined by whether the AI was able to understand the context deeply enough to trust you.

How Our Technical Setup Influenced Artificial Intelligence

Key Takeaway: Clear context within the code convinced the algorithm to recommend us, even though we weren't in the top spot in traditional SEO. The AI overview appears above organic results. If you are featured in it, you are more visible than anyone else below you.

How to Get AI to Cite You

Writing high-quality content is a prerequisite today; you can't build authority without it. However, in the era of artificial intelligence, it is no longer enough. Today, algorithms need to see clear relationships and a logical structure that they can process in a flash.

Good content is just the beginning. AI needs to look under the hood.

In practice, this means that marketing can no longer operate in a silo, separate from the technical team. It’s not just about what is written on the website, but how its "chassis" (the underlying architecture and back-end data) is built. That is exactly where it is decided whether the AI will understand you and ultimately recommend you.

Artificial intelligence loves facts, not marketing fluff. To get it to choose you as a source, follow these three principles:

• Speak to the point: AI prefers clear definitions and simple sentences without unnecessary adjectives. If you make a claim, state it directly.
• Sentence independence: Try to write so that key sentences make sense even on their own. These punchy insights are exactly what AI loves to "pull" directly into its responses.
• Consistency is key: If you demonstrate your expertise consistently over the long term – not just in articles, but also in your website's technical code, other professional portals, and social media – the algorithm will start to perceive you as a true authority in your field.

Simply put: The less effort the AI has to put into understanding your website, the more likely it is to recommend you to customers.

The Technical Background: How a Machine Sees Your Company

To make the AI's job easier, we don't just rely on standard text; we also use structured data (in JSON-LD format). This is a special layer of code that acts as your company’s and services' "ID card," written in a language that algorithms understand instantly and without errors.

Example: This is how we define a company and its specific service within the code:

Why Code Alone Is Not Enough

A code entry tells machines what you do, but on its own, it won't force them to actively start recommending you. AI needs context. It wants to know exactly who stands behind your service, what your long-term track record is, and how this information fits with the rest of your website. It is precisely this depth of interconnectedness that determines whether you earn the algorithm's trust.

Key Takeaway: Structured data (JSON-LD) is essential, but it is not enough on its own. AI needs to see consistent expertise – who stands behind the service, how long they have been doing it, and where else it is mentioned. Without this depth, code alone will not save you.

How We Implement GEO for Our Clients

Getting a company into AI overviews is not about adding a single line of code. We build an entire "data layer" that provides algorithms with absolute certainty.

• The Expertise Map: First, we identify the specific areas where you want to be perceived as an authority. We don't just look for what people are "googling" today; we focus on where your true strengths lie.
• Translating into Machine Language: We then translate this expertise into semantic code. We interconnect information about your organization, specific services, and professional outputs. As a result, the algorithm knows exactly who provides the service and who is signed under it as an authority.
• Consistency: Your expertise must be equally readable in your content, within your code, and across sources outside your website (such as LinkedIn). The more this information aligns, the less the AI doubts you.

A Technical Foundation That Builds Trust

Loading speed, clean code, and precise data measurement (server-side tracking) are not just technical terms to us. They represent the vital infrastructure that provides algorithms with a sense of stability and reliability. This is exactly how digital visibility transforms into real inquiries and actual business.

Key Takeaway: GEO is not a one-line code fix. It is built across three layers simultaneously – the expertise map (where you want to be an authority), the translation into machine language (semantic code), and consistency across your website and external sources. The less the AI has to guess, the more likely it is to recommend you.

Discover more about our comprehensive approach to SEO, GEO, and marketing.

Checklist: Test How AI Sees Your Company

Instead of just reading the theory, try it out for yourself. These questions will reveal whether your website is readable for modern algorithms:

1. Run a quick test in ChatGPT or Google Gemini: Ask: "Who are the best experts in [your core service] in Slovakia?" or "Which companies would you recommend for [your line of business]?" If the AI fails to mention you, it means it lacks sufficient, reliable data about who you are.
2. Do you have a "digital ID card" in your code? If you looked under the hood of your website, would you find clear definitions of your company, services, and experts (structured data)? Without them, the AI is forced to guess. Ask your developer if your website utilizes JSON-LD schemas.
3. Is your website technically fit? Does your website run lightning-fast and error-free? If bots struggle with speed or accessing your content, the AI will simply skip your site.
4. Do you say the same thing about yourself everywhere? When the AI compares your website with your LinkedIn profile or industry portals, does it see consistent expertise, or does it find conflicting information?
5. Are you measuring data in a "cookieless" era? Can you reliably track the customer journey and the success of your marketing even under today's strict privacy rules?

What We Don't Know Yet: Open Questions in GEO

Let’s be honest: GEO is a field that changes month by month. The algorithms behind AI overviews are proprietary, documentation is scarce, and most of what we know about GEO today comes from rigorous testing and benchmarking. At iMPROVE, we are actively tracking several questions for which definitive answers do not yet exist:

How long does it take for AI to start recommending a company post-implementation? We see the effects within 2 to 3 weeks for some clients, while for others, it takes several months. We cannot yet identify the exact differentiator with absolute certainty, though we hypothesize that the density of external mentions prior to the technical overhaul plays a critical role.

Which AI platforms respond most strongly to JSON-LD? Google AI Overviews and Perplexity clearly parse structured data. ChatGPT and Claude behave less predictably; their citations often suggest they draw more heavily from external mentions than from the code on your website.

Does GEO work the same way in Slovak as it does in English? Our tests indicate that it does, but with one major caveat. In English, AI models have access to data spanning a much longer historical period. Slovak websites only began to be properly and meaningfully indexed around 2024, when the processing of Slovak natural language improved significantly. For Slovak companies, this means everyone is practically starting from scratch. Whoever begins building consistent signals today stands to gain a first-mover advantage that will be incredibly difficult for competitors to close down.

Where is the "consistency threshold" that triggers an algorithmic recommendation? In other words, how many external mentions, reviews, profiles, and citations are required before an AI decides, "I know this company well enough to feature them"? There is likely no single magic number, but rather a gradient. We are currently working to map this out.

How does GEO evolve with each new model iteration? As models update, their preferred sources can shift. This is exactly why we don't focus on short-lived hacks that work today but break tomorrow. We anchor our strategy on principles that remain stable across model versions: clear expertise, consistent signals, and a technically clean website. Those will never go out of style.

These questions don't just stay theoretical. Every project teaches us something new – and as we uncover definitive answers, we plan to share them right here on our blog.

Conclusion: If the AI Can't See You, You Don't Exist to the Client

The way businesses source vendors has fundamentally changed. Your potential clients are no longer spending hours sifting through pages of Google links. They are asking artificial intelligence.

Today's rules are unforgiving: you are either part of the AI's response, or you are invisible to the client.

The companies that successfully bridge the gap between marketing and technical precision today will be the ones the AI recommends as the top choice tomorrow. At iMPROVE, we know how to "translate" your website into a language that AI understands, building the kind of authority that algorithms simply cannot ignore.

Frequently Asked Questions (FAQ)

1. What exactly is GEO, and how does it differ from traditional SEO?
SEO (Search Engine Optimization) focuses on getting your website to appear in Google's list of blue links. GEO (Generative Engine Optimization) goes a step further – it optimizes your website so that artificial intelligence (such as ChatGPT or Google Gemini) can comprehend, process, and directly recommend you within its generated response.

2. Do I need to rewrite my entire website for AI?
Not at all. GEO is not about deleting your old content; it is about enhancing it. The core requirement is to add a technical layer to your source code and refine your messaging to ensure it is clear, factual, and stripped of unnecessary marketing fluff. AI heavily prioritizes hard facts over empty adjectives.

3. How soon will I see the first results of GEO optimization?
AI models are constantly learning and re-crawling the web. The initial shifts in how AI perceives and cites your brand can become noticeable within a few weeks after implementing the technical code updates and semantically interconnecting your data.

4. Is GEO suitable for smaller B2B companies?
Yes, and it often represents a massive opportunity for them. In traditional SEO, smaller companies frequently face an uphill battle against corporate giants with massive backlink profiles. In GEO, however, success isn't determined solely by the size of your budget, but by the clarity and verifiability of your expertise. If you are an expert in a niche field, GEO allows the AI to find and recommend you over a large competitor with an ambiguous website.

5. Can a standard webmaster handle these technical updates?
GEO requires tight collaboration between a marketer and an experienced developer. It is never just about "pasting a piece of code"; it demands a deep understanding of data semantics and web architecture. This is exactly why we approach GEO comprehensively – handling everything from strategy and expertise mapping to the final technical implementation.