GEO optimisation, because AI recommends the brands it knows.

GEO optimisation (generative engine optimisation) prepares your website, content and brand data so that AI systems such as ChatGPT, Perplexity or Gemini know and name your company. The aim is a place in the fully written answer rather than a place in the list of results, including in Google’s AI summaries. The work has three levers: technical access for AI crawlers, content that answers customer questions directly, and mentions on sources the systems trust. GEO complements classic search engine optimisation and builds on it.

SEO optimises for rankings in the results lists of Google and Bing, GEO for mentions in the answers of AI systems such as ChatGPT or Perplexity. With SEO, people choose from ten results themselves. With GEO, the AI has already made the choice and usually names only a handful of providers. That is why measurement differs: instead of positions, what counts is the share of answers in which your brand appears. Both share the same foundation, namely technically clean and credible content. Good work on the website almost always pays into both.

AI visibility describes how often and how accurately AI assistants such as ChatGPT, Perplexity or Google’s AI Overviews name, describe and link to your brand. There is no position number as on Google: either you appear in the answer, or a competitor takes your place. Because every answer turns out slightly differently, AI visibility is measured as a hit rate across many questions and repetitions. The how counts too. A mention with the wrong location or an outdated offer does more harm than good.

The cost of GEO optimisation depends on the number of questions measured, the languages and markets, the state of your website and the amount of new content. An AI visibility audit is a clearly defined starting point: it shows your baseline and the biggest levers. Ongoing support covers content, mentions and the monthly measurement. Support starts with a three-month minimum term and is then cancellable monthly. After the initial call, you get a fixed-price quote with no hidden items.

At Maze & Friends, GEO optimisation runs in six steps: strategy call, baseline measurement, technical foundation, answer content, mentions and monthly re-measurement. In the strategy call, we define the questions where your name should come up. The baseline documents, across at least 20 purchase-decision questions, how AI systems describe you today and whom they name instead. Then we put crawler access, indexing and company details in order before any new content is created. You approve every new page before it goes live.

A GEO agency makes sure AI systems know your brand, describe it correctly and name it for the right questions, and it measures whether that works. At Maze & Friends, this includes an AI visibility audit, the technical foundation with crawler access and structured data, answer content for your website and mentions on sources already cited on your topic. Every month we re-measure with the same questions. Some providers call the same work AI SEO or LLM optimisation (LLMO).

You measure AI visibility by regularly asking several AI systems a fixed set of customer questions and logging whether your brand appears. Because answers fluctuate, ask each question several times and record hit rates rather than single results. Four figures are telling: mentions, links, your share compared with competitors and the sources that own the topic. GA4 also shows visits from AI assistants, and server logs show hits from AI crawlers. There is no counterpart to Google Search Console for ChatGPT.

No, nobody can seriously guarantee a recommendation by ChatGPT, because AI systems decide for themselves which sources to name, and their answers fluctuate. What we commit to is work on every known factor: technical readability, citable content and mentions on trustworthy sources. On top of that comes a monthly measurement with the same questions, which shows you whether the work is having an effect. We don’t use hidden instructions to AI systems in page copy, even if they might bring mentions in the short term.

GEO optimisation works step by step: technical improvements usually take effect first, while mentions in AI answers build up over several months. Systems with live search, such as Perplexity, react faster to new content than the trained knowledge of the models, which only changes with new model versions. AI crawlers don’t visit every day either. For highly contested questions it takes longer, because established providers have been present there for years. That is why we compare the same questions month by month and assess trends rather than individual answers.

Yes, SEO still makes sense, because AI systems draw on search indexes for current answers and Google remains the most used search engine. Google’s AI Overviews feed directly on the Google index, and what isn’t crawlable and indexed can’t be found or cited by ChatGPT’s web search either. GEO therefore builds on a working SEO foundation and adds a layer on top. If the foundation is shaky, we tell you and start with classic search engine optimisation.

AEO (answer engine optimisation) shapes content so that specific questions are answered with your answer, for example in featured snippets, voice assistants or AI answers. GEO targets generative systems such as ChatGPT and Perplexity, which rephrase answers from several sources. In practice, the two overlap heavily. The same tools help: questions as headings, a complete answer in the first sentence and paragraphs that make sense on their own. LLMO and AI SEO essentially describe the same field of work.

Optimising a website for GEO means making pages readable for AI crawlers, company details unambiguous and content a direct answer to customer questions. Technically, that means not blocking the AI providers’ search crawlers, such as OAI-SearchBot or PerplexityBot, delivering content without JavaScript and maintaining structured data based on Schema.org. On the content side, clear definitions, verified figures and paragraphs that stand on their own count. You usually don’t need a new website for this; the existing one can be retrofitted. Keyword density plays no role.

You get into ChatGPT’s answers when ChatGPT knows your brand from its training or finds it through web search, and that is exactly what SEO for ChatGPT aims at. For web search, the OAI-SearchBot crawler must be allowed to read your pages, and what gets cited are mainly trustworthy pages that answer a question clearly, plus directories, trade media and comparison articles that mention you. For questions about agencies in Austria, ChatGPT and Perplexity mainly named service pages and industry directories in our measurement from October 2026. A mention can’t be forced, but it can be prepared.

You get into Google’s AI Overviews through the Google index: only pages that Google crawls and indexes can appear there as a source. AI Overviews are the AI summaries above the search results. Pages outside the top ten results are cited too, if they answer a question precisely. What helps is the answer in the first sentence, questions as subheadings, tables for comparisons and videos, because YouTube is among the most cited sources. Classic SEO is the entry ticket here.

An llms.txt file does no harm, but its effect on visibility in ChatGPT, Perplexity or Google is not proven. The file summarises a website’s most important pages for language models. No major AI provider has committed to reading it on third-party websites, and analyses of crawler traffic show hardly any requests from the relevant bots. It is mainly useful for developer tools that read software documentation. We set one up because the effort is small, but we don’t sell it as a measure.

Our GEO optimisation covers ChatGPT, Perplexity, Google AI Overviews, Google Gemini and Claude, and we measure in all five. Microsoft Copilot can’t currently be evaluated automatically, so we spot-check it by hand and say so openly. The basic principles apply to every system. In the detail they differ, for example in how heavily they rely on live search or trained knowledge and which sources they prefer. Perplexity names considerably more sources per answer than ChatGPT and is therefore often the first place where progress becomes visible.

GEO mainly brings mentions and trust, with traffic from ChatGPT only in second place, because many people don’t click on the sources. The effect often shows later: someone who came across you in an answer searches for your name on Google or gets in touch directly. That is why, besides visits from ChatGPT or Perplexity in GA4, we also measure mentions and brand searches. We don’t promise a fixed increase in traffic. But if you are missing from the answer, you never even get compared.

A reputable GEO agency measures your starting point before selling you services and doesn’t guarantee mentions in ChatGPT. Ask which questions, systems and repetitions are used for measuring and whether you get the raw data. Make sure crawler access and indexing are checked before any new copy is written. Avoid providers who write hidden instructions to AI systems into page copy or buy reviews. Tricks like that come out, and the brand pays the price. A good sign is an agency that measures and publishes its own AI visibility.

For GEO in Austria, partly different sources count than in Germany, so the Austrian market is measured from an Austrian location with its own questions. In our measurement from October 2026, questions about agencies in Austria surfaced sources such as the Austrian Economic Chamber’s (WKO) Firmen A-Z company directory, wlw.at, domestic trade media and press releases via the APA-OTS news service. A mention in Berlin therefore says little about Vienna. Maze & Friends from Vienna measures Austria, Germany and Switzerland separately.

Yes, you can do the basics of GEO yourself: collect customer questions, answer them directly on your website and make your company details consistent everywhere. The first test is simple too. Ask ChatGPT, Perplexity and Google the questions where you want to be named, and note who gets named instead. Clean measurement across many questions and systems, structured data and building mentions on third-party sources take more effort. For that, a partner who re-measures the same questions every month pays off.