How does one of the world’s largest insurance brands adapt to AI Search?
We sat down with Daniel Reeh, Head of Digital Marketing at Zurich Insurance (Germany), who leads website, performance marketing, social media, content – and now AI search – in one team. He also teaches generative engine optimization (GEO) at TH Köln.
His short version of the shift: “Same same, but different.” The long version is worth your time.
AI is becoming the new gatekeeper
Daniel doesn’t sugarcoat it: website traffic is declining because of AI.
“People are searching or chatting with Gemini, ChatGPT, whatever. AI will maybe be the new gatekeeper in the future.”
And he pushes back on the idea of “classical” search as a separate channel:
“What is classical web search? Google has AI Overviews, AI Mode — everything is shifting to AI. It’s crazy to talk about classical old search engines.”
In Germany, the engines that matter for Zurich right now: ChatGPT, Gemini, and Perplexity.
Less traffic, but it converts remarkably well
The trade-off Daniel sees: fewer visits, higher quality.
“When there’s traffic on the website, it’s highly conversional. You see very high conversion shares of relevant, quality traffic.”
The catch is measurement. The customer journey now happens inside AI chats — a black box:
“The journey is more in a big box now. You only see the results in the end, but you lose the journey in the middle.”
For Zurich, that end result matters more than raw traffic anyway. As an agency-based (not direct) insurer, the website’s job is to generate leads and connect the online journey to face-to-face conversations in local agencies.
Brand mentions are the new standard
Asked what replaces rankings as the core metric, Daniel is clear:
“In the new AI search world, brand mentions will be the new standard.”
Mentions and citations feed each other — citations are often the result of brand mentions across the web. Which changes where the work happens:
Reputation management beyond your own website. What are people saying about you elsewhere? Product tests, reviews, Wikipedia, Wikidata.
You can’t optimize one page anymore. “The AI system has a whole overview of the website. You have to see the whole thing and invest more in your brand.”
Knowledge graphs are back — and bigger. Google’s Knowledge Graph dates to 2012, but with GraphRAG behind modern AI systems, structured entity data is more relevant than ever.
Tell the AI your USPs — explicitly
One practical takeaway: don’t assume AI systems know what makes you different.
“We are focusing on telling the artificial intelligence: these are our unique selling points, this is our strategy. It’s not like everybody knows it — you have to implement those things.”
That means putting your positioning in content and metadata (e.g., schema/knowledge graph markup), so AI systems can understand your website and brand.
There are no proven best practices yet
Daniel also teaches AI search at TH Köln — and warns his students against anyone selling certainty:
“It’s a new field. There are not really proven best practices. People are trying a lot.”
His example: llms.txt files, which many rushed to implement before Google and other providers said they’re not relevant. His advice instead: transfer what worked in SEO, build a genuinely good product and USP, invest in reputation — and use monitoring tools to see what’s actually working for your brand.
Key takeaways
AI is the new gatekeeper — the journey happens inside ChatGPT, Gemini, and Perplexity before anyone reaches your site.
Expect less traffic, higher conversion — measure quality and outcomes, not visits.
Brand mentions are the new ranking signal — and drive citations.
Reputation management > page optimization — reviews, product tests, Wikipedia, Wikidata all feed AI answers.
Make USPs machine-readable — state them explicitly in content and metadata.
Be skeptical of “best practices” — the field is too new; test and monitor instead.
