Jan-Willem tells us which metrics really matter in his view, why you should look beyond impressions, and how to build a consistent brand that AI can recognize and cite
Search is changing faster than most business websites can keep up with.
AI Overviews, AI Mode, ChatGPT, Gemini, Claude, Perplexity: every week brings a new acronym, a new “foolproof” checklist, a new metric to track.
How much of all this actually helps those who have to grow a business every single day? That is exactly the question we want to answer.
In SEO Confidential we interview the world’s top SEO and GEO experts to bring you concrete, useful answers designed for your daily work. The goal is to help you understand what really matters to become the source chosen by AI-powered answer engines.
The guest of this new episode is Jan-Willem Bobbink of the agency 11 Internet, a specialist in enterprise SEO, international and local SEO, large-scale website management and custom tool development.
He built and ranked his first website in 1997, ran his first e-commerce store in 2004, and today speaks at international events such as SMX Munich, BrightonSEO and SEOCampixx. His philosophy rests on three pillars: more leads, higher conversions, better user experience.
In this interview Jan-Willem gave us a practical guide, full of insights and useful points of view, on what to actually do to build an authoritative brand in the era of answer engines.
If, as I suspect, your goal is not vanity metrics but leads and conversions, you absolutely cannot miss this conversation.
Make the web describe you consistently
There is a lot of talk these days about more transparent ranking criteria and rules designed to make search fairer. From your point of view, what would really change for a business if Google were forced to explain how it favors certain content (and brands) over others? Do you think this would make it easier to build a brand that gets chosen and cited by AI as well?
Less than most people hope. Transparency helps you measure results, but measuring is not the same as having control. We are already getting a taste of it.
On June 3, 2026 Google added a report dedicated to generative artificial intelligence to Search Console, which finally shows how often pages appear in AI Overviews and AI Mode. It is worth remembering that this new feature arrived in the wake of regulatory pressure from the United Kingdom, not through Google’s own initiative. And the first thing you notice is that the report only shows impressions: no clicks, no CTR, no information about queries. In other words, even this newly won transparency only tells you that you were visible, without telling you whether that visibility produced any concrete value.
If Google were forced tomorrow to make the entire logic of its algorithm public, the main effect would probably be a new arms race among those who already know how to optimize very well. Being cited by AI systems is a consequence of being a genuinely authoritative source that is clearly identifiable on a given topic. Knowing the recipe does not automatically make your content the best answer, nor does it turn your brand into the entity the model trusts.
Transparency makes the scoreboard readable. It does not score the points for you. To a business owner I would say: wish for better measurement tools, but do not fool yourself into thinking that access to the algorithm’s details can build your brand’s strength.
Google is starting to show how often a site appears in AI Overviews and the other AI-powered search experiences, but it still provides no data on the clicks generated by these citations. From your point of view, which metrics should a business owner track today to understand whether their brand’s presence in AI answers is producing real business value?
Treat the new AI impressions report as an indicator of visibility, not value, because that is exactly what it is. The number to watch is not the absolute figure but its trend. Is the share of impressions coming from AI growing week after week? And is it growing faster than impressions from traditional organic search? If the answer is yes, it means answer engines are relying on your content more and more.
To measure real business value, on the other hand, you need to cross-reference several indicators, because Google deliberately chose not to show click data. I monitor branded search volume as an indicator of the awareness generated by AI, because people who see your brand cited in an answer often search directly for the company name instead of clicking on the source. I also watch direct traffic and so-called dark traffic, GA4’s AI Assistant channel to capture the clicks Search Console does not show, assisted conversions and, above all, qualified leads and revenue generated by high-intent sessions.
The share of citations earned on Bing can also be useful as a trend check. Finally, I always check the server logs to see which AI crawlers are actually visiting the site, because they are a direct, independent data source that no dashboard can alter.
In short, the raw number of citations is a vanity metric. The real question is a different one: is being cited increasing brand awareness and generating new business opportunities? The answer lies in your analytics data and your CRM, not in an impressions chart.
Bing Webmaster Tools shows which topics a brand gets cited for by AI systems. How reliable is this data? What mistakes might a business owner make if they read it as a faithful snapshot of how AI perceives their brand?
The Topics view was released in preview on June 16, 2026, together with the Intents, Citation Share and Compare features. I am glad it exists. What convinces me less is how it will be interpreted. Microsoft itself is very cautious: it describes this data as observational metrics, explains that the classification is still evolving, and notes that for niche sites the labels can be rather generic. That is a warning worth taking seriously, even though many will probably ignore it.
The first mistake is confusing a citation log with a perception model. Topics shows the topics you have been cited for. It does not tell you what the model thinks of your brand. Those are two completely different concepts.
The second mistake is treating it as a complete picture. Bing and Copilot represent only part of the artificial intelligence ecosystem and, for many sites, most of their exposure in AI systems happens on platforms Bing cannot observe.
The third mistake is reorienting your entire editorial strategy on the basis of a classifier that is still in preview and bound to change over time. It is also worth remembering that grounding queries are machine-generated reformulations, not the queries users actually typed.
For this reason, Topics is useful for spotting gaps and opportunities in your content coverage, and that is precisely its strength. It should not, however, be presented to a client as the tool that shows “how AI sees your brand”, because that is not its purpose and Microsoft has never claimed anything of the sort.
If an AI system associates a brand with the wrong topic, what consequences can that have on visibility and reputation? Are there concrete strategies to help models better understand a company’s real identity?
This is an entity problem, and it is one of the few GEO issues that causes real trouble. If a model classifies you under the wrong topic, you show up for searches that do not convert while failing to appear for the ones that do, and you silently lose market share to a competitor whose identity is unambiguous. The worst case is conflation, where a similarly named entity overlaps with your profile and you inherit associations that do not belong to you. That is a reputation problem, not just a visibility one.
The solution lies in corpus-level consistency, not in clever markup. Models learn your identity from everything that has been written about you, so you need to make that body of text internally consistent.
Put your own site in order first, with a clear “About” page, consistent naming and “sameAs” links to your real profiles. Where you are relevant enough, make sure Wikidata and Wikipedia are correct, since they are the reference point for much of entity understanding. Then make your footprint on third-party sites consistent, meaning your mentions on Reddit, YouTube, review sites and the trade press, because that is exactly what models actually read. Schema markup reinforces all of this, but it is a supporting signal, not the main lever.
The principle is simple. You cannot trick a model into understanding you. You have to make the open web describe you consistently, and the model will follow.
In recent months dozens of GEO checklists have appeared, promising to increase your chances of being cited by AI. Which pieces of advice are actually backed by data, and which risk being nothing but fads with no concrete results?
The list backed by solid data is short and boring, and that is exactly the point. Make sure you are indexable, because if indexing bots cannot find and render you, nothing else matters. Write clear, direct answers with a clean, easy-to-extract structure. Keep your entity consistent. Build an authentic presence on third-party platforms, where the data is more solid.
A meta-analysis pooling more than fifty studies found that, when it comes to AI citations, brand mentions outweigh backlinks by roughly three to one, and that cited content tends to be significantly fresher than uncited content. Freshness and being talked about are real signals. Real authority and real expertise are real.
The list of fads is longer and noisier. llms.txt, sold as a citation lever, is the prime example because crawler logs simply do not confirm it. Keyword stuffing in prompts, “magic content chunking” and AI-specific rewriting have all been explicitly flagged by Google as unnecessary practices. Obsessive AI-oriented schema falls into the same category. The pattern to distrust is anything packaged as a brand-new acronym with a checklist attached. There is credible research showing that tactics that sound convincing often fail to beat an unmodified baseline in a controlled test. When a tactic cannot beat doing nothing, it is a fad, not a method.
If you had to give a single piece of advice to a business owner who wants to be chosen as a source by ChatGPT, Gemini or AI Overviews, what would you focus on? Do you think it makes sense to talk about GEO techniques, or is it just a fashionable acronym?
One piece of advice. Become the clearest, most genuinely useful and most cited source on your specific topic, and do not block the crawlers that could cite you. That is the whole secret condensed into a single sentence.
Is GEO a buzzword? Largely yes, and I say that as someone who works in it. The term repackages the fundamental principles that good SEO and PR professionals have been putting into practice for years.
What is genuinely new is specific and concrete. Measurement is new, citation mechanisms are new and crawler access has become a real-time decision rather than an afterthought. So I use the word reluctantly. If I had to quantify it, I would say the work is about ninety percent quality content, authority and technical correctness, and about ten percent things genuinely specific to AI-powered answer engines. Anyone offering you the opposite ratio is selling you the buzzword, not the actual work.
How much do structured data and Schema.org really affect the probability of being cited by ChatGPT, Gemini or AI Overviews? Are they a decisive factor or just one of many signals taken into account?
It is not decisive. It is just one signal among many, and a supporting one at that. Google has stated clearly in its AI optimization guidelines that structured data is not required for generative AI search and that there is no specific schema to add for that purpose. It still recommends using schema, but for rich result eligibility within regular SEO, not as a “trick” to earn citations.
Schema’s real value lies in disambiguation and automated parsing. It helps a system understand what an entity is, how much a product costs, who an author is. That contributes to entity clarity, which I care about a great deal. But content quality, crawlability and brand authority do the heavy lifting and far outweigh schema. My rule of thumb is that schema is the finishing touch that makes good content machine-readable. It does not turn weak content into a cited source, and you should not invest in it expecting a citation boost from it alone.
If you were auditing a site that wants to become a go-to source for AI answer engines, where would you start? Structured data, entity organization, content quality or something else?
I start with the least glamorous item on the list, namely accessibility. Can AI crawlers reach you and, once they get there, can they render what matters?
First I check the server logs to see who is actually fetching the site, then the robots.txt file and rendering, because a site that delivers its content via client-side JavaScript can rank well for human users and be nearly invisible to an indexing bot. If GPTBot, OAI-SearchBot or PerplexityBot cannot read your site properly, every other improvement is just window dressing.
After that, the order is: content quality and direct-answer structure, then entity organization, and finally structured data as reinforcement. So the sequence is: access, then authority, then identity, then markup. People tend to start with schema because it feels like a concrete checkbox. Schema is the final layer, not the foundation, and checking it first is a good way to polish a site that search engines cannot read properly in the first place.
Blocking AI crawlers protects your content but reduces your visibility in answer engines. If you were running the site of an SMB that lives on authority and customer acquisition, would you let all crawlers in, block them, or make a platform-by-platform selection?
Allow selectively, leaning heavily toward “allow”. For an SMB whose growth depends on being found and trusted, a blanket block is usually counterproductive. You are trading content protection for invisibility on the very platforms your future buyers are starting to use. That is a bad trade for a small business that needs every last bit of visibility it can get.
The nuance to understand is that not all bots do the same job. Retrieval and citation bots, the ones that gather information to build an answer and link back to your site, are the ones you want to let in, because blocking them directly costs you citations.
Crawlers dedicated exclusively to scraping for training purposes are a more reasonable place to draw a line if you have genuine intellectual property concerns, because their value to you is less direct. So make a platform-specific decision, document it, and revisit it as the landscape changes.
One specific warning: the new Google Search Console opt-out button that removes you from AI Overviews and AI Mode is something almost no SMB should touch. You would give up AI visibility without gaining any ranking benefit.
On the llms.txt file, some consider it a fad and others an emerging standard for the agentic web. Is it worth implementing right away, or are there higher-priority tasks to focus on first?
For most businesses it is a secondary priority. I am not against it. If your CMS gives it to you for free and it takes ten minutes, publish it and move on, because the downsides are practically nil. What I object to is the idea that it earns you citations today, because crawler data says otherwise. Ahrefs examined 137,000 domains and found that roughly 97% of published llms.txt files were never requested in May 2026. Other large-scale log studies reach the same conclusion. The bots that generate citations mostly ignore the file.
Its actual use is narrow and specific. It is a business-to-agent routing surface, particularly valuable for developer documentation and for the tools that AI coding agents read directly. That is a fact, and that is exactly where I would expect it to grow in importance as the agentic web develops.
For a typical SMB marketing site, content quality, crawlability, entity clarity and third-party presence all far outweigh it. So my position is: keep an eye on it, do not bet on it, and never let it push the fundamentals aside. It solves the capability-discovery problem for agents, which is a different matter from getting your answer cited.
On one hand Google says llms.txt is useless for SEO, on the other it uses it on its own sites and integrates it into Lighthouse. How do you explain this contradiction, and what should a business owner conclude from it?
It looks like hypocrisy, but it is not. These are two product teams answering two different questions. The Search team has been consistent and direct. John Mueller compared the llms.txt file to the old keywords meta tag, and Gary Illyes confirmed that Google is not taking it into account. Their point is that it is not a signal for Search, for AI Overviews or for AI Mode, all of which draw from the same index.
On May 7, 2026, the Chrome team released Lighthouse 13.3 and added an Agentic Browsing check to the default set; that check verifies whether the file exists. This is about browser agent compatibility, not ranking. Same company, two different domains, no actual conflict once you understand the connection.
Google publishing llms.txt and markdown in its developer documentation follows the same reasoning. Mueller explained it as a token-saving convenience for the AI coding agents that read the reference material, not as an SEO move.
So the takeaway for a business owner is clear. Do not read Google’s use of the file as an endorsement for your marketing site, because the context is completely different. It may matter for agents in the future. Right now it is not a lever for Search, and anyone telling you to read the Chrome audit as a ranking signal is misunderstanding which team is speaking.
The difference between showing up and actually mattering
This interview with Jan-Willem Bobbink confirms something worth remembering every day: the race for acronyms comes and goes, while the fundamentals always stay the same. Clear content, real authority and a consistent brand identity are worth far more than any temporary trick.
The point I find most valuable is precisely the distinction between visibility and value. Knowing how many times you appear in an AI answer is only a starting point. The real question is about the leads, revenue and conversions that visibility actually generates. That is why it becomes essential to look beyond the dashboards and cross-reference the data with analytics and your CRM.
Equally illuminating is his reasoning on crawlers: blocking everything out of fear means giving up precious opportunities just as customers are starting to use these tools to look for solutions. A selective, informed choice looks decidedly wiser, and I have no doubts about that, as I wrote here.
A sincere thank you to Jan-Willem for the generosity and clarity with which he shared his experience. I am sure his advice, the fruit of his work and of real case studies, can be genuinely useful to you if you care about growing your brand.
See you next week with another international guest here on SEO Confidential.
#avantitutta
