Picture Martina, who is opening a café in Bratislava and needs a new website. A few years ago she would have typed “web design Bratislava” into Google and clicked through the first three pages of results. Today she opens ChatGPT and writes: “I’m looking for an agency to build a website for my café, who would you recommend?”
Within seconds she gets a few names, a short rationale and links to sources. She doesn’t click anywhere, she simply picks from what the AI offered her. If your company isn’t in that answer, you don’t exist for Martina. And that is exactly what all the fuss around the acronyms SEO, AEO and GEO is about.
Why this matters right now
Just a few years ago, search was simple: you type a query, Google returns a list of links, you click. Today Google shows AI Overviews and AI Mode, ChatGPT has its own search, Perplexity builds its entire product on answers with sources, and Microsoft Copilot does the same. You get the answer ready-made, often without having to go anywhere.
In February 2024, Gartner predicted that the volume of traditional search would drop by 25 % by 2026, precisely in favor of AI assistants. We treat that as a forecast, not an established fact, but the direction is obvious, and visits from chatgpt.com or perplexity.ai now routinely show up in website analytics. For a business, this boils down to a simple question: when someone asks an AI for a supplier in your field, will it mention you? And can you influence that at all?
You can. It isn’t magic, it’s mechanics. And that is exactly what we’ll take apart here, both in plain terms and under the hood.
How AI search works
In plain terms
Classic Google is a librarian. You tell it a topic, it points you to a shelf of books sorted by which ones it considers best, and you leaf through them yourself. An AI search engine is more like an advisor: it listens to your question, runs into the library itself, skims a few books, notes down the essentials and says: “According to these three sources, here is how it is.”
That advisor doesn’t make things up; most of the answer is assembled from what it found on the web. So the question is: are your “books” in the library, can they be opened quickly, and is the answer written right at the start of the chapter?
Under the hood
A language model on its own doesn’t know the current web; its knowledge ends at its training date. That is why systems like Google AI Overviews and AI Mode, ChatGPT search, Perplexity or Microsoft Copilot work on the RAG principle (retrieval-augmented generation). They first break your query into several sub-searches, fetch the relevant pages, pick specific passages from them, and only then does the model assemble an answer with citations from those passages.
The key word is “passages”. The system doesn’t cite your whole website but a specific paragraph or block that directly answers a sub-query. Two completely different things therefore come into play: whether the search engine finds you at all (classic SEO) and whether your text is written so that it can be extracted and quoted (that is the new discipline). The model prefers sections that answer directly, are structured, factual, up to date, come from trustworthy sources and are consistent with what other websites say about the topic.
SEO, AEO and GEO: what’s what
In plain terms
SEO is when you try to be the first book on the shelf. AEO is when you want the librarian to read one sentence straight from your book as the answer, without the customer having to open it. GEO is when you want the advisor to mention you by name and point to you as a source while assembling their answer.
All three rest on the same foundation: an accessible, fast and clearly written website with truthful content and a trustworthy brand. The difference lies in what you count as the result: a click, an answer or a mention.
Under the hood
- SEO (Search Engine Optimization) is optimization for classic search results. It rests on content relevance, the technical health of the website (speed, indexing, mobile), backlinks and the E-E-A-T principle: experience, expertise, authoritativeness, trustworthiness. It is measured in rankings and traffic.
- AEO (Answer Engine Optimization) aims to get the system to pick you as the direct answer: in Google’s featured snippets, in voice assistants and in AI answers. It works with concise answers to specific questions, FAQ sections and a clean structure where the answer sits at the start of the paragraph.
- GEO (Generative Engine Optimization) is the youngest term. It was coined by a 2023 academic paper titled “GEO: Generative Engine Optimization” by authors from Princeton, Georgia Tech, IIT Delhi and the Allen Institute. The goal is for generative search engines to cite or mention your brand and content in their answer. It isn’t about a ranking, but about whether you are part of the synthesis.
In practice, these disciplines overlap far more than three acronyms would suggest. Most of the work you do for GEO is good SEO at the same time. The reverse doesn’t necessarily hold: a website that sits on Google’s first page thanks to backlinks but has its answers buried in long marketing paragraphs is something AI often skips.
What decides whether AI recommends you
In plain terms
Back to the advisor in the library. For them to recommend you, three things must be true. They must be able to reach your book, meaning it isn’t locked away in storage. They must find the answer in it quickly, not on page 47 under the heading “Our philosophy”.
And they must trust you. It works like a recommendation from friends: if you ask five people about a good dentist and three say the same name, you believe them. If other books say something different about you, or don’t mention you at all, the advisor would rather recommend someone they know from several sources.
Under the hood
The first prerequisite is accessibility for crawlers. If you want to be visible in these systems, your robots.txt must not block GPTBot, PerplexityBot or ClaudeBot. Mind one nuance: Google-Extended controls the use of your content for Gemini, but AI Overviews draw on the regular Googlebot. Blocking Google-Extended therefore won’t make you disappear from AI Overviews.
The second prerequisite is readability. Key content must be present in the HTML even without JavaScript running, because many crawlers don’t render the page, they only read the code. The website must be fast, headings should say what a section is about, and the answer should sit at its start. schema.org structured data (Organization, Article, FAQPage, Product, Service) are machine-readable labels that tell the system what is a company, what is an article and what is a question with an answer.
The third prerequisite is trust and consistency. Your company name, address, prices and claims should be the same on your website, in your Google Business Profile, in directories and in articles others write about you. If your website shows different opening hours than the map does, the system isn’t sure and would rather reach for a source that doesn’t contradict itself. Also worth a mention is llms.txt, a proposed standard from 2024 by Answer.AI meant to help models find their way around a website. It is cheap to add, but there is no guarantee that every system uses it.
Advantages over classic SEO, and what is worse
Let’s start with what is better. A visitor who reaches you from an AI answer is usually further along in their decision. Martina from the opening is no longer figuring out what a website is; she is comparing specific agencies. Intent is high, and such a person is closer to a decision than someone who clicked through from the tenth link.
Second advantage: your brand shows up even without a click. When an AI says “I recommend company X”, it works like a recommendation from a friend, not like an ad. The third advantage is temporary: competition is still lower, few companies tackle this systematically. And the fourth is practical: work on content, structure, technology and authority pays off in SEO and in AI alike. You aren’t doing two things, you are doing one thing properly.
Now the downsides. Zero-click is a reality: if the AI answers on your behalf, many people never reach your website, even when it mentions you. Measurement is harder, because AI answers aren’t stable rankings; today ChatGPT mentions you, tomorrow, with the question phrased differently, it mentions a competitor. And you depend on changes at platforms that have promised you nothing and owe you nothing.
Is it worth it?
The honest answer: it depends on who you are and what you sell. Yes, if your customers ask questions and compare before buying (“who can fix…”, “what is the difference between…”, “how much does… cost”): services, B2B, specialized online stores, consulting, healthcare, finance, trades. Yes also if you already have a decent website and content and just need to restructure it so that AI can read from it.
Not yet, if your website lacks the basics: it is slow, isn’t indexed or has no content. AI optimization would be a roof without walls; classic SEO has to come first. Be careful, too, if you live off traffic itself (media, on-site advertising), where zero-click may take away more than it adds, and with purely impulse purchases, where nobody asks anything before buying.
Realistic expectations: we don’t promise rankings or growth percentages; nobody honest can today. What we can do is prepare your website so that the system can read it, understand it and quote it. Whether it does also depends on what the rest of the internet says about you. The surest return is a side effect: a website that is also better for regular search and for people.
How to measure it
There are no rankings in the classic sense here, so you measure differently. In Google Analytics 4, filter referrals from domains like chatgpt.com or perplexity.ai, create a separate segment for them and track how their numbers develop and, above all, their conversions: inquiries, orders, bookings. The numbers are usually smaller than from Google; the trend matters more.
The second pillar is tracking brand mentions. Put together a set of 20 to 50 questions your customers might ask and enter them regularly into ChatGPT, Perplexity, Copilot and Google AI Mode. Record whether you are mentioned, in what context, whether with a link, and who is mentioned instead of you. You can do this by hand or with tools; what matters is regularity and the same set of questions.
The third pillar is the ratio of your mentions to your competitors’, which follows directly from that set; it is the most understandable indicator of progress. Search Console won’t help you with AI directly, as AI Overviews aren’t reported separately in it. Use it rather to check that your key pages are indexed and free of technical errors.
In closing
SEO isn’t dead, it just got a more demanding reader. One that no longer scans a list of links but picks passages, compares sources and assembles an answer. The good news is that people, Google and this advisor all want essentially the same thing: clear answers, true facts and a website that opens and reads effortlessly. Whoever starts today is still playing on a less crowded field.
If you aren’t sure where your website or campaigns stand, we’ll gladly take a look. We audit websites and campaigns free of charge and with no strings attached; just get in touch. We’ll tell you frankly what works, what doesn’t and whether GEO is even worth your attention right now.