A few months ago, a prospect told me they had found me through ChatGPT. They had typed something like "performance marketing expert in Greece" and my name came up in the response. They did not find me through Google Search. They did not click a paid ad. They asked an AI, and the AI recommended me. I want to explain why that happened, because it did not happen by accident, and if you work in B2B, the same dynamic is starting to affect whether your clients find you or your competitors.
How LLMs Decide Who to Recommend
A common misconception persists that LLMs work like a search engine with better language. They do not. A search engine retrieves pages that match your query and ranks them by authority and relevance. A language model generates an answer based on patterns in its training data, and when it has access to real-time search (like ChatGPT with Search, Perplexity, or Google AI Overview), it also retrieves current sources to supplement that answer.
This means visibility in AI answers depends on two distinct mechanisms that require different strategies. The first is training data presence: whether you existed as a recognized entity in the web content the model was trained on. The second is real-time retrieval: whether you appear in the sources the model's search integration finds when it looks for current information to support its answer.
SEO affects the second mechanism more than the first. A high-ranking page can get retrieved by an LLM's search integration when relevant. But entity recognition, the AI "knowing" who you are and including you naturally in responses, comes primarily from training data. An entity that appears consistently, is cited by credible sources, and has clear structured information attached to it tends to appear in answers even without real-time retrieval. An entity that exists only as a well-optimized landing page gets overlooked when the model generates from its base knowledge.
Entity Recognition: Making Sure the AI Knows Who You Are
An entity, in the context of how LLMs understand the world, is a distinguishable person, organization, or concept with consistent attributes attached to it across multiple sources. "George Tsiros, Performance Marketing Director in Athens" is an entity when those three facts appear together consistently across LinkedIn, your website, publications you have written for, and anywhere else your name appears professionally.
Inconsistency breaks entity recognition. If your LinkedIn title says "Digital Marketing Expert," your website bio says "Performance Marketer," and your byline in a guest post says "Marketing Consultant," the model has difficulty determining that these are the same person with a clear professional identity. The more consistently your name is paired with your specific expertise and location across sources, the more confidently an LLM can include you in answers about your field.
Entity building is straightforward but requires consistent execution. Your LinkedIn profile, website bio, guest post bylines, directory listings, and any speaking or event appearances should all use exactly the same name, title, and location. Schema.org Person markup on your website, with name, jobTitle, url, and sameAs links pointing to your LinkedIn profile and any other active profiles, gives LLMs a machine-readable identity card for who you are. This markup does not train the model, but it does help the model correctly extract and associate facts when it retrieves your page.
Citation Authority: Be in the Sources LLMs Trust
When LLMs retrieve real-time sources, they weight heavily toward publications and domains they have seen cited frequently in their training data. For performance marketing, that means industry publications, well-linked blog posts, agency directories like Clutch and G2, and professional platforms like LinkedIn. Being mentioned, not just ranked, in these sources is what creates citation authority.
A single well-placed byline in a credible industry publication is worth more for LLM visibility than ten optimized landing pages on your own site. When a respected publication publishes an article with your name in the byline and your expertise as the subject, that publication's authority transfers to your entity. LLMs that retrieve that article as a source are simultaneously exposed to your name in the context of your expertise, exactly the pattern that builds the association you want.
Being quoted in journalism is even stronger. When a journalist cites you as a source in an article, "according to George Tsiros, Performance Marketing Director at Advengers", that attribution creates a high-authority citation that LLMs treat as a signal of recognized expertise. The structure is not unlike what SEO has always called link building, but the mechanism is different: citation authority and entity association, not PageRank.
For B2B professionals specifically, appearing on lists, "top performance marketing professionals in Greece," "leading digital marketing directors to follow", creates the kind of categorical citation that helps LLMs answer queries like "who are the best performance marketing experts in Greece?" These lists are indexed, they are credible, and they explicitly associate your name with a category. Getting on two or three credible lists in your specialty is more impactful than extensive on-site content optimization.
What Content Formats LLMs Actually Cite
Not all content is equally likely to be retrieved and cited by an LLM. The formats that appear most consistently in LLM responses share a few characteristics: they are definitional, they are specific, they contain verifiable facts or numbers, and they are structured in a way that makes them easy to parse.
Definitional content, "What is Performance Max?", "What is GEO?", "What is Marketing Efficiency Ratio?", matches directly to the query patterns LLMs are most frequently asked. If your site has a clear, well-written answer to a question someone is likely to ask an AI, and that answer is indexed and accessible, it has a realistic chance of being retrieved as a source.
List-based content performs well for similar reasons. "5 reasons your Google Ads ROAS is wrong" or "The 7 things to check before launching Performance Max" maps to how LLMs structure answers when asked for recommendations or steps. They tend to retrieve and integrate list-based content naturally because it fits the format they default to when explaining multi-step topics.
Comparison content, "Performance Max vs Smart Shopping," "Google Analytics 4 vs Universal Analytics", serves decision-making queries that buyers actively ask. When a CMO asks an AI "should we use Performance Max or keep our Shopping campaigns?", the AI will retrieve comparison content if it is available and credible. Creating clear comparison content on questions your prospects genuinely face positions you as a source for exactly the questions they are taking to AI before they talk to you.
The single most important characteristic is specificity. LLMs strongly prefer citing content that contains numbers, percentages, and concrete claims over vague qualitative statements. "PMax CPA typically improves 8 to 20 percent after the learning phase" is citable. "PMax can improve your performance" is not. Every piece of content you produce for GEO purposes should contain specific, verifiable claims your audience can act on.
The Exact Queries Your B2B Clients Are Asking AI Right Now
The most useful thing you can do before designing a GEO strategy is spend 20 minutes running the queries your ideal clients actually ask. Open ChatGPT, Perplexity, and Google AI Overview separately and search these:
- "Who are the best performance marketing agencies in Greece?"
- "What should I look for when hiring a performance marketing director?"
- "How much should I spend on Google Ads for a mid-size e-commerce business?"
- "What's the difference between Performance Max and regular Google Ads?"
- "How do I know if my Meta Ads are actually working?"
- "What is GEO and how is it different from SEO?"
For each query, note: who appears in the answer, which sources are cited, and what claims are made. That exercise tells you your exact competitive landscape in AI search, who has GEO visibility today, which questions are being answered confidently versus hedged, and where there is a gap between the quality of existing answers and what you could provide. The gaps are your content opportunities.
LinkedIn: The Disproportionately Powerful B2B GEO Signal
LinkedIn content receives disproportionate citation weight from LLMs for two reasons. First, LinkedIn's domain authority is exceptionally high, it is one of the most linked-to sites on the web, and LLMs trained on web data have seen LinkedIn content cited extensively. Second, LinkedIn explicitly associates professional identity with content: your name, title, company, and location are attached to everything you publish there. This is exactly the entity-and-expertise association that LLM citation systems value.
The implication for B2B GEO is that consistent, substantive LinkedIn publishing is one of the highest-leverage activities available. A post that generates genuine professional engagement, comments from practitioners, shares by people in adjacent fields, responses that continue the conversation, creates a signal that is indexed, attributed to your professional identity, and associated with your area of expertise. It does not need to go viral. It needs to be substantive enough that other professionals engage with the content rather than just scrolling past.
Publishing long-form articles on LinkedIn (not just posts) creates additional indexable content on a domain LLMs cite heavily. These articles should address questions your target clients are asking, the same questions you identified in the exercise above, and they should contain the specific claims and numbers that make content citable. A LinkedIn article titled "What a Performance Marketing Director Actually Looks at Before Recommending a Budget" that contains real decision-making frameworks and specific criteria will attract more professional engagement and more LLM citation weight than a general article about marketing trends.
Structured Data as LLM Scaffolding
Structured data in Schema.org format does not directly train language models. But it does make it significantly easier for LLMs to correctly extract and associate facts when they retrieve your pages as sources. Think of it as scaffolding: it helps the model understand what the page is about, who wrote it, when it was published, and what specific claims it makes, all of which affect how confidently the model cites the content.
For a personal professional site like this one, the highest-priority schema implementations are Person markup on your main page (with name, jobTitle, url, sameAs pointing to LinkedIn), BlogPosting markup on every article (with author, datePublished, headline, description, and wordCount), and FAQ markup on service or expertise pages where you answer specific client questions. FAQ schema in particular maps directly to the question-answer pattern that LLMs use when generating responses, a well-structured FAQ on your site is essentially giving LLMs pre-formatted content in exactly the form they prefer to cite.
How to Measure GEO Progress
GEO visibility is harder to measure than SEO rankings, but it is measurable. The most direct method is also the simplest: run your target queries in ChatGPT, Perplexity, and Google AI Overview monthly and record who appears and what sources are cited. Do this consistently and you will see movement over three to six months. The queries you identified in the client-perspective exercise above are your benchmark set.
Watch for your domain appearing in Perplexity's cited sources. Perplexity explicitly shows the pages it retrieved to formulate its answer, which makes it the most transparent of the major AI search systems for GEO auditing. If your content starts appearing as a cited source for queries in your area, you have genuine GEO visibility, not just content that exists, but content that the AI is actively using when it generates answers.
GA4 will start showing perplexity.ai and chatgpt.com as referral sources as AI-driven traffic grows. These are still small numbers for most accounts, but they are growing quarter-over-quarter, and the quality of leads from AI referral tends to be high: someone who found you through an AI recommendation has already been pre-qualified by the AI's answer before they arrived at your site.
The B2B GEO Flywheel: Where to Start This Week
GEO compounds in a way that paid media does not. The content you publish today trains future model versions. The citations you earn this year become training data for models that launch next year. The entity recognition you build through consistent professional presence is self-reinforcing, the more sources associate your name with your expertise, the more confidently LLMs include you in answers, which leads to more branded searches, which strengthens entity recognition further.
The starting point this week is not complicated. Run the query exercise above and document your current baseline, where do you appear, where are your competitors, and what are the content gaps. Then identify two or three specific questions your target clients are asking AI that you can answer better than any existing source. Write that content with the specificity and structure that makes it citable. Publish it on your site with proper schema markup and amplify it on LinkedIn. Reach out to one industry publication about a guest contribution that addresses a question in your target query set.
None of these actions individually will produce overnight visibility. Collectively, sustained over six to twelve months, they build the kind of entity recognition and citation authority that makes you the answer when your next client asks an AI who to call.