This is how your company will be recommended by ChatGPT – the GEO guide

Imagine a buyer looking for a new solution. Three years ago, they would have googled, opened five vendor websites, compared prices and references, and built a longlist. Today, they are increasingly asking ChatGPT or Perplexity the same question – and instead of ten links, they get a ready-made answer with three or four named vendors.

For you, this means: the pre-selection takes place before anyone has seen your website. Those who appear in the AI's answer are in the running. Those who don't are simply non-existent to that decision-maker at that moment – no matter how well your own website ranks.

The shift is measurable. In the first four months of 2026, 68.01 percent of US Google searches on desktop and mobile web ended without a click on a result (SparkToro using clickstream data from Similarweb, June 2026). Two years prior, the same figure was 60.45 percent. So, the trend is clearly heading in one direction. And on our form fields, „ChatGPT" now regularly appears as the answer to the question of how interested parties became aware of Lilox.

The good news: AI visibility is not a matter of chance, but a system – and most of your competitors are still ignoring it. AI systems preferentially cite sources they already know and consider reliable; those who get in early will continue to be cited. A self-reinforcing effect that will be difficult to catch up with later.

Read on AI visibility is one of several building blocks. The guide shows how it fits into the overall system. B2B Lead Generation.

Before you optimise anything, you need a baseline. The good news: the first check costs nothing but half an hour. Simply ask your AI systems the questions your customers would ask – and document who is mentioned.

For this to be meaningful, you need three types of questions. Category Queries see if you even appear in the relevant set: „Which B2B lead generation agencies in the DACH region can you recommend?" Problems Requests to show if the AI connects you with the problems that you solve: „Our rankings are stable, but organic traffic is falling – who can help?" And Brand inquiries to see if what the AI tells you about them is true at all: „What do you know about [your company]? What does it stand for?"

Here are the prompts for each in ChatGPT and Perplexity: **ChatGPT:** * "Translate the following text into English (UK)." **Perplexity:** * "Translate the following text into English (UK)." and Gemini – the three systems draw from different sources and often give quite different answers. For each answer, note: Are you named? In which position? With what reasoning? And at least as importantly: Who is named instead?

Repeats the same question every month, word for word. Only the time series turns a snapshot into a measuring instrument – and later shows you in black and white whether the measures are working.

If a decision-maker asks AI for a solution like yours and you aren't mentioned in the answer, it almost always has one of four causes. The good news: each of them can be fixed.

Building Block 1 · Answer Format

AI systems cite what they can cleanly extract: clear definitions, concrete figures, FAQs in the verbatim wording of real search queries. A text that needs three paragraphs to get to the point is difficult for a language model to cite. A sentence like „GEO refers to the optimisation of content for generative AI systems," on the other hand, is a perfect answer capsule. By the way, be cautious with superlatives: hype language tends to lower the citation rate rather than increase it.

Building Block 2 · Entity and Schema

The AI must understand in a machine-readable way: "who" you are, "what" you offer and "where" It's where you're active. Structured data ensure this: Organisation, WebSite, and FAQPage schemas, plus a canonical who-what-where statement that appears consistently everywhere – website, LinkedIn, directories. Inconsistent details measurably confuse language models.

Building Block 3 · External Mentions

AI systems are distrustful of self-reported information. Their recommendations are largely fuelled by what Third Writing about yourselves: Industry directories, review platforms, expert articles, LinkedIn articles. Those who only exist on their own website remain an unconfirmed rumour to AI.

Block 4 · Open Crawlers

The most banal and common mistake: AI crawlers are locked out. GPTBot, ClaudeBot, and PerplexityBot respect robots.txt – if there's a Disallow, they simply cannot read your content. This also includes llms.txt, a machine-readable business card at the root of your website. You can check if your crawlers are open in one minute at yourdomain.com/robots.txt.

Error 1: Blocking AI crawlers to „protect" content. A understandable reflex, but counterproductive for most B2B companies. Those who block do not appear in AI responses – and leave recommendations to the competition.

Error 2: Optimise once, measure never again. GEO is not a project with an end date, but a loop. AI responses change with every model update; without monthly measurement, you won't notice any progress or regressions.

Error 3: Producing definition articles with no added value. AI has long been able to answer interchangeable „What is X" articles itself. Those who can prove expertise, provide their own data, and are suitable as a source are cited – case studies with figures, documented processes, real experience.

Regarding expectations: First mentions usually arise within four to eight weeks, first around the brand name and on niche topics. Category queries like „best solution for..." take longer and, above all, more external signals – there you compete with everyone who is already being cited.

Read on Three building blocks, one system – LinkedIn lead generation · Marketing funding

This reflects the real prompts from ChatGPT, Perplexity & Google AI concerning AI visibility, GEO and AEO.