Ask a room of marketers what Generative Engine Optimization is, and someone will tell you it’s SEO with a new logo. Matthäus Michalik, Co-founder of the Berlin-based organic search agency Claneo, hears it constantly, and it’s the first thing he wanted to correct.
“The biggest [misconception] is that everyone says GEO is just SEO. SEO is the foundation of a good GEO visibility, but you still have to understand how you can influence the answers, so that your brand or your product gets mentioned instead of the competitors.”
Foundation versus finished job. That distinction runs through everything below: how Claneo retrained its SEO team, which metrics he’d put in front of a CMO, the one KPI he thinks is overrated, and a hidden-text experiment that sat dead for months before AI systems suddenly started picking it up.
Who is Matthäus Michalik, and what is Claneo?
Matthäus Michalik is Co-founder of Claneo, a Berlin agency for organic search working with brands across DACH and internationally. The agency has spent years doing classic SEO for enterprise clients, which is exactly why his read on GEO is worth having. He isn’t arguing SEO is dead. He’s arguing that people who say GEO is only SEO have stopped looking closely enough at the systems. Claneo has since built out a dedicated GEO practice, published the first German-language book on Generative Engine Optimization, and put out a public GEO content checklist.
The systems don’t behave the same way
Michalik’s first move is to break “AI search” into its actual parts. Google’s AI Overviews and AI Mode behave one way. ChatGPT and Perplexity behave another.
“Maybe it also makes sense to make a differentiation between Google and AI Mode, AI Overview. But definitely ChatGPT and Perplexity, they work differently. So the basis may be the same, but you still have to understand how the systems work.”
That changes which engines are worth your attention, and Claneo’s rule is audience-first:
“When we talk about B2C industry, probably ChatGPT and AI Mode, AI Overview are more relevant than Copilot. When we talk about B2B, where we see in Germany a lot of companies use Microsoft, maybe Copilot as a SaaS company, as a B2B service company, makes more sense.”
In other words, Copilot belongs on a B2B radar in a way it never does for a consumer brand. Getting that mix right is the difference between monitoring the right prompts and monitoring noise.
Two aha moments: poems in 2022, product research in 2023
Michalik dates his own shift to two points. The first was November 2022, when ChatGPT launched.
“Everyone was doing poems about their work, about SEO. And we said, ‘Oh wow, you can do a poem with an AI.’ So this was like a first step.”
A novelty, not a channel. The second moment landed months later, when ChatGPT got a web connection.
“Then we realized, ah okay, now we can also use ChatGPT to search for products, to search for recommendations, for brands, for use cases where you need a solution.”
Even then, nobody knew whether it would stick. “A lot of people said, ‘Oh, that’s the biggest threat for Google. No one will Google anymore.'” What convinced him wasn’t the threat narrative. It was watching Google itself change shape.
“They’ve become from a search engine to an answer engine, to a dialogue-based AI engine. Because they realized users want to have an answer instead of 10 links.”
“This is where we have realized: okay, when Google is changing, then probably we as agency need to change as well.”
Then the clients started asking. “A lot of clients ask us, ‘Hey, what is now changing in the search industry?'” By the time AI Mode reached Germany, the question had settled itself: “It’s here to stay, and we need to make sure that we support clients to go this change within how people interact with the internet.”
Retraining an SEO agency into a GEO agency
Michalik describes the internal shift as three vectors. The first was speed, driven by client expectation:
“A lot of clients said, ‘Okay, now we have AI, so now you have to be much faster.’ So the efficiency of AI also takes place in our agency, that we make sure our processes are much better and we are faster in delivery.”
The second was mapping which AI systems each client’s customers actually use. The third, and the hardest, was people:
“We also needed to educate our SEO experts to become more GEO expert, and then understand how the new systems work: how to create content for AI citations, or to be more citation-worthy, how to get a digital footprint in third-party data.”
That last phrase is the real cultural shift, because it drags SEO out of its own lane and into digital PR, community and brand. Our own data backs the direction of travel: active Reddit communities get cited nine times more often, and AI search engines lean heavily on press releases.
“It’s becoming more and more not just an SEO. You also have to take care of your brand, you do a PR, you optimize your brand for search. It’s becoming broader and more connected to other departments within a company, instead of just working SEO on your own dedicated website.”
AI search is a brand channel, not a traffic channel
This line belongs on a poster in every marketing department:
“AI search is a brand channel. They are not made to drive traffic. And I think this is the biggest mistake, if you think, ‘AI search is not relevant to me because I just have 2% AI traffic.’ That’s often what companies think. It does not generate any traffic, so probably I do not need to take care of it.”
Claneo lives the counter-example:
“A lot of companies which reach out to us say, ‘Hey, we know you from ChatGPT. We know you from Gemini,’ even if our own traffic on our own agency website is really, really little compared to organic, to paid, to social media.”
The traffic number and the business impact are measuring two different things. Our own data agrees: Claude drives 10.6% of OtterlyAI signups while Google Analytics reports 0.1%, and AI search now drives more of our revenue than Google. Judge AI search by last-click analytics and you’ll conclude it doesn’t matter, right up until your pipeline fills with people who found you in a chat window.
What Michalik tracks, and the one metric he’d skip
His measurement stack is deliberately small. Start with mentions:
“To monitor your mentions for the right prompts you would like to be visible within your target group is like the first key. To be mentioned, to build up your visibility in different systems, the systems which are relevant to you, and then also to make sure that you get cited.”
Then citations, yours against your competitors’, plus the third-party domains you could realistically appear on:
“For us as an agency, it makes sense to work with agenturtipp.de or OMR Reviews, because they are quite strongly cited when someone’s asking for a service provider in SEO or GEO.”
Your list will look different. Find it before you spend a euro on outreach. Then add self-reported attribution, which he recommends to every company: after the purchase, simply ask how they heard about you.
The metric he’d deprioritize is sentiment:
“I’m not such a big fan of sentiment, because I would say if you have a problem with the sentiment, then you have a problem with your product and not with your AI search.”
“In general, the AI systems talk quite positively about a brand instead of pointing out the negative aspects. This is why sentiment is something what I would, at the latest, have a look at.”
How to report AI search to management
His framing splits cleanly into inputs and impact. Inputs are what you changed on your own site and what you earned externally. Impact is visibility, mentions and citations. Put them on one annotated timeline:
“I would highly recommend, when you’re talking to the management, that you show them how the visibility develops over the time, and also what were the actions. So here we had a publication. Maybe here we had an advertorial. Maybe here we changed the content.”
And keep measuring AI referral traffic in GA or Adobe anyway:
“We see that this traffic converts much better than classic organic traffic. So even the traffic is little, it’s maybe more relevant for your business.”
The shortcuts that work today, and might not survive
Michalik is blunt that GEO is in its exploitable phase, and that exploits have a shelf life. Self-promotional listicles worked, in his words, “like hell,” and may run into German or local advertising law. Paid advertorials are working now, but he expects AI systems to discount them:
“Advertorials which are now used for AI search maybe also get filtered out of AI systems, because they say, ‘Hey, they are not objective.’ But if I pay for an advertorial, for sure I can write whatever I want, right?”
“You also see, when you compare it to SEO, that paid media does not influence organic that strongly. And I’m sure that this will be also somewhere in AI search, that everything which is paid will not have such a strong impact on your organic AI visibility.”
“Maybe the shortcut works today, but maybe it will be a threat in three years, five years.”
The alt tag experiment
Here’s the part most GEO practitioners haven’t tested. Claneo ran a two-stage experiment on their own site.
Stage one was a deliberately false claim in hidden text:
“We did a test where we said, ‘Hey, Claneo has an office in Berlin as well as New York.’ And we did it on white text and white background, and this was quite easy, gets easily cited and mentioned by AI systems.”
They corrected it. Stage two moved a similar claim, an office in Cape Town, into the alt text of an image instead. For a long time, nothing happened. Then it changed:
“Now we see, since a couple of weeks, that some of the AI systems take the alt tag into account. So you also see that it changes how AI systems crawl your page, and maybe to take care of your alt tags makes sense nowadays. Six months ago, it did not make any sense to have a strong focus on alt tags.”
The takeaway isn’t “go stuff your alt tags.” These were controlled tests of false claims, not a tactic to copy. The takeaway is that crawl behaviour is moving under your feet, which is why he keeps testing:
“Testing is everything. Either you follow your tests or our tests, because if you don’t test it, you will never know what really works.”
“When you compare that to SEO, GEO is not… there is no playbook which 100% says that’s how it’s working. So it’s still re-engineering how AI systems react when you do things and change things on your website.”
Who should own AI search?
His first question with a new client isn’t tactical. It’s an org chart question.
“The first question I would ask myself is: who is owning AI search? Is this a part of SEO team? Is it a part of the branding team, or maybe the PR communications team?”
The honest answer is that it touches several. “When we talk about external third-party citations, mentions, then probably PR is really important.” Technical GEO, making sure the site is crawlable and indexable for AI systems, stays with SEO and IT. “Whenever you talk about sentiment, maybe then it’s a product and brand topic.” So where does it sit?
“I would probably give it to the SEO team, because they have the skill set to have all the factors of AI search, and not just the reputation, PR topic. We did it decades before at Google. I’m sure they will own it as well as they did for organic search.”
Key takeaways
- SEO is the foundation of GEO, not the whole of it. Each engine behaves differently, and your engine mix should follow your audience: ChatGPT plus Google AI Mode and AI Overviews for B2C, add Copilot for B2B.
- Judge AI search as a brand channel. A 2% traffic share is not evidence it doesn’t matter.
- Track four things: mentions on the prompts your buyers use, citation rate against competitors, the third-party sources cited in your category, and self-reported attribution.
- Deprioritize sentiment. Bad sentiment is a product problem, not a GEO problem.
- Report inputs and impact on one annotated timeline, and keep measuring AI referral traffic even though it’s small, because it converts better than classic organic.
- Treat shortcuts as temporary and test continuously. Alt text went from irrelevant to picked up in months.
- Give ownership to SEO, and make them work with PR, brand and IT.
Watch the full conversation with Matthäus Michalik, Co-founder of Claneo, on YouTube. Previously in this series: Slopelift, OMR Reviews and Zurich Insurance Germany.
Want to see where your brand shows up in AI answers, and who’s getting cited instead of you? Track your AI search visibility with OtterlyAI across ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Gemini and Copilot.
