How to Choose an Agency to Get Your Brand Cited by ChatGPT

… So multiple reasons. We normally run a 29-step health check to see why you’re not being cited. But very high level, you always want to start with, is your website actually technical accessible for the AI crawlers? For example, if you have a lot of JavaScript, that’s not good because, for example, ChatGPT does not read JavaScript, so it can only read the meta description and meta title of your website, but nothing else.

… Very often, for example, when you host a website on Cloudflare, it per default blocks GPT bot, Perplexity bot, Claude bot, and so on. Okay so you have to proactively enable it in your robots.txt file and put allow so that the bots can actually go and read your content on your website.

… Three steps how you can find out the buying intent or like, the whole intent across the customer journey. Social listening. You go on Reddit, Quora, YouTube comments, TikTok, Instagram, … and scrape these platforms for the questions that people ask there around your topic along the customer journey. And then you find the seed keywords and find more and more long-tail keywords. … Second one is go and ask your customer success and sales team, what are the most frequently asked questions that we getting? … And then the third one is that you extract your Google Search Console, GA4, and Bing Webmaster tool data.

Because LLMs search context-based. Every one of us has their own context in their LLM. So you use ChatGPT, it saves the context about you, about your work, who you are, where you lived, where you live, what you have, what kind of preferences you have, and matches this when doing a search for you, what matches you best. So when you look for a product, we both look for a TV, will both of us give us different answers, and that has to do with the intent because the intent for you to buy a TV is different than for me when I buy a TV.

SEO is optimizing for also technical readability of your website and optimizing to rank high for certain keywords. And AEO is really understanding the intent of your customer that can be super different and based on the persona and the location and the context within the LLM and all the search engines work differently. ChatGPT has its own index. Perplexity searches on the web. Claude is searching through Brave. And in traditional SEO, you were mostly targeting only Google and only focusing on one algorithm. Now you’re focusing a lot of different algorithms and a lot of different personas.

That’s absolutely correct. … Yes, you can. Like, uh, doing good SEO is already 70% of it. … And AEO then is, if you now again look at larger enterprises, is something that is cross-functional. It goes across different divisions. … you have to start talking to customer success, to sales. And you have then to talk to product to understand how to answer these questions. With SEO, it was a lot easier because you were mostly focused on, like, the technical optimization. … Now, you have to talk to everyone in the organization to actually understand and orchestrate the efforts to become visible for a lot of different intents.

JavaScript, they can’t parse, read JavaScript, so they just won’t ignore it. They will ignore it. … And for example, um, funny, um, case, like in the, in the hotel and travel, um, industry, a lot of these widgets that they’re using where you can book hotels or book flights, um, and compare prices, these widgets are very often rendered in JavaScript, and the problem is that the crawlers can’t access them. So very easy growth hack to become more visible when you’re travel agency, for example, have them in clean HTML and, um, the AI will have a higher probability of actually giving you as the answer when someone is looking for a price.

Funny thing, YouTube is now, according to a lot of, uh, researchers, one of the most cited sources in AI overviews on Google. But the LLMs can’t watch the YouTube videos. So what people now do is they publish a YouTube video … and the video description is very well optimized to answer … the question in the title of the YouTube video, which is then being pulled into the AI … overview.

So I think it was Opus 4.8, um, where it started taking screenshots from websites to look at the, look at the page to actually see what i- what is on there.

You always want to have the subline describing what is in the picture. … You always want to keep the effort for the LLM as low as possible because they don’t want to spend a lot of tokens. That’s why it’s so important that also everything is in clean HTML because they don’t have to spend a lot of tokens.

Yeah. So for example, there’s this rule that you, um, that, that’s a rule from SEO, how long should your meta title, meta description be so that it’s being shown in the SERP results, so in the normal, uh, blue link results, and that it’s fully readable. Now with AI also, um, you want to have a H1 that has a length of 60, um, capitals. Then you want to have a subheading with like 150 characters that summarizes what the, the ar- article is answering. And then you want to have a chunk with key takeaways, which is summaring, summarizing important points of the whole article so that no one has to actually read the article.

The human reader not. Because no, the, the readability for when you are, um, chunking too much into context, we get very, we got very bad feedback that was when that was done too much because it’s just not readable anymore.

… What’s even more interesting is what does AI-ready context look like to be turned into content? Because the big question or the, the… for example now here in Germany, these called Mittelstand companies, … they make up ninety-five percent of our GDP. … These companies have a tremendous amount of unique knowledge that made them compete on a global level for a long time, and they’re these hidden champions. … How would they actually become the answer in AI search? It would be super easy for them because they have all of this unique value and knowledge that they accumulated over decades no other company in the world has. The problem is this is stored in … spreadsheets, old books, databases that are inaccessible. … The question should not be is your content AI ready, but is your knowledge or context AI ready?

Because we surpassed the point where there is more AI-written content on the web than human-written content. And at the same time, the foundational model companies have crawled the whole web and have trained their models on the whole web, and they are licking their fingers for this implicit expert knowledge that hasn’t been written down yet. So if you’re an expert in a field that is writing down his or her knowledge, and the AI knows and can verify that it’s actually you, for example, because you link your LinkedIn profile, you have an author page that is well-structured, AI will be way, way more likely to actually cite you.

Keep it as raw as possible because that will make it unique.

… if let’s, let’s, let’s take the example. So you and Ringmaster, you build a content cluster that is called, um, um, what AI needs to know about B2B marketing, and then you have the sub-cluster GEO And then you have and then you have a blog article interview with, with Simon. This is the main cornerstone piece, and then you can have a lot of blog posts which all only go around this one question that you gave me. And there you can do a lot of research around it and then use a quote, like a real quote from this transcript in the blog article from me citing me. … This will make the, the value of the content even better than just like purely writing down my answer, but using it in the contextual better just like right way.

… I want to really understand why do you want to present for this prompt? How did you test it? And then we test it and assess if you’re really absent. We then … assess is your page actually technically optimized so that if we start publishing content that it actually works. If not, we will have to start with technical optimization. If technical visibility, we call it technical visibility, is given, readability is given, then we go ahead and have a catalog of questions that you have to go through where you provide us … what you do, how do you define your focus, which market, which product do you focus on. We normally always start with one market, one product, set up the knowledge base around it. Once all of this is set up, we then define the 30 to 50 prompts, super transactional bottom of the funnel based on the three pillars that I talked earlier about social listening, internal data, customer success, sales, and all the data that we can get through your Google Search Console, Bing Webmaster Tools, and GA4. Then we track are you visible, are you not visible, who is visible, and why are they visible? Lay out the color clusters, subclusters, produce the first three articles. They get these articles. They have to human review first the outline and then the article itself and give feedback. This is important because we don’t publish fully AI-generated content. There’s always a human in the loop, but the human in the loop needs to be at the best point in time so that it’s very time efficient for them because no one at these large enterprises has time. And then we publish the content.

… The whole setup of the brand kit, the knowledge base is the most manual and time-consuming part on our side because it’s super hard or it’s nearly impossible to collect all of this knowledge and context without having a human going to them, talking to them. Or you just do a podcast and you have all of these questions that they have to go through either way and ask them these questions on the spot. And then in the end you give them, “Hey, these were the answers. Are they correct? Can we use them for the content creation?”

Prompt tracking is a scam. … No one from all the big players — Profound, PKI, RankScaile, whatever — have found out the exact way to do it that they can literally academically prove and tell you, “You are now the answer.” Because LLMs are not deterministic. So how do you actually assess if it’s working? How much more revenue have I made by people finding me in AI? That’s the only valid metric. You can in Google Analytics look at the clicks that are coming directly through the AI chatbots or AI overviews, and can then follow them to have they booked a meeting or have they bought the product. Second one is self-attribution. In B2B, you can in your book a meeting form, just ask them, “Where did you find us?” B2C, D2C, include in your email to the customer a short form, where did you find us? Because this email has a very, very high open rate because everyone wants to know when their package is arriving.

Have they driven revenue through AI search visibility?

How do you measure success? How much more revenue are we making for being visible in AI search? … 60% of the searches already end without a click because of an AI summary. If we’re not investing into this channel, we’re going to be invisible or absent even though we might have great brand reputation, but the next generation of people who are going to buy us will not even know that we exist even though we poured millions into brand campaigns.

… the capacity to do GEO because GEO is cross-functional and to actually become visible, you don’t have to only measure who’s visible for prompts, but you actually have to execute and publish great content. And to do this in large organizations is a pain.