Expert Take — Hans Schepers
As long as advertisers pay more for worse converting visitors, organic wins — even though that game has become more complicated than ever.
Last updated: April 2026
Why “optimise for AI” is too vague
I see it in nearly every GEO article: “Make sure you’re visible to AI platforms.” Fine advice, but it’s missing something fundamental. Because when you look at which sources those platforms actually cite, you see three completely different patterns.
From recent citation research:
- ChatGPT pulls 47.9% of its top citations from Wikipedia and encyclopaedic sources
- Perplexity pulls 46.7% from Reddit
- Google AI Overviews leans 23.3% on YouTube and multimedia content
And here’s where it gets interesting: only 11% of domains are cited by both ChatGPT and Perplexity. What you do for one platform largely doesn’t work for the other.
That changed how I think about GEO strategy. “Optimise for AI” is too generic. The question is: which AI platform, for which type of query, via which retrieval mechanism?
Do you still need traditional SEO?
This surprised me: around 90% of ChatGPT citations come from pages ranked position 21 or lower in Google’s traditional search results. A top-3 Google ranking is not a prerequisite for LLM citation. Good news, really 😉
That’s counter-intuitive if you’ve spent years investing in SEO positions. But it makes sense when you realise that LLMs combine training data with selective browsing — and training data has a much longer time horizon than last week’s SERP.
Does that mean SEO becomes irrelevant? No. Google SEO rankings still correlate 0.65 with LLM visibility. It helps indirectly. But it’s no longer the only route — and for many queries, not even the most important one.
What actually counts?
Two things stand out in the data.
Content structure. 44.2% of all LLM citations reference content from the first 30% of a text. Front-loaded answers — where the core claim or conclusion comes first, before context and nuance — get cited significantly more often. That’s the exact opposite of how much B2B content is written, where the conclusion only arrives after three paragraphs of context.
I find that a valuable lesson. Not just for GEO, but for everything you write. Start with your point — that’s what I do here with a TL;DR (Too long, didn’t read).
Brand search volume. This correlates more strongly with LLM citations (0.334) than backlinks. People actively searching for your brand name — that’s a stronger signal for AI platforms than the number of links pointing to your site.
Consistency across multiple sources
There’s another pattern I keep seeing in the data: AI systems gain more confidence in a recommendation when it appears consistently across multiple independent sources. Not just your own website. Not just one review. But a coherent presence in relevant Reddit discussions, YouTube videos, trade publications, review platforms, and your own content.
The system is looking for confirmation from multiple angles. That’s a different investment than just optimising your own site.
What I take from this
I’m still actively experimenting with this, but my preliminary conclusions:
- Stop treating “optimise for AI” as generic advice. Decide which platforms are relevant for your audience and adapt your approach per platform. I’m actively researching this myself and will bring you along.
- Reconsider your content structure. Put your conclusion at the top — not at the bottom.
- Brand awareness is (probably) more important for GEO than backlinks. That’s a fundamentally different investment logic than what SEO has been built on for the past decade.
- Consistency across multiple sources matters more than dominance on one platform.
None of these points are definitive answers — the data is fresh and the landscape is changing fast. But the direction is clear enough to adjust your strategy around.
Sources: LLM citation benchmark research — Profound (2025), Semrush platform citation analysis, Princeton/Georgia Tech GEO paper (KDD 2024)
FAQ
Do I need a separate strategy for each AI platform?
Not necessarily separate, but you need to know which platforms your audience uses and understand that the sources they cite are fundamentally different.
Are backlinks no longer important?
They help indirectly via SEO rankings (correlation 0.65 with LLM visibility). But as a direct factor for AI citations, they count less than brand search volume.
What do you mean by “front-loaded content”?
Put your conclusion or main point in the first 30% of your text. Not as clickbait, but as a clear answer. AI platforms cite that first 30% disproportionately often.
How do you build brand search volume?
Not with one tactic. It’s the sum of consistent communication, visibility across multiple platforms, and content that gives people a reason to remember your name and search for it.
Which AI platform matters most for B2B?
It depends on your market. But for B2B in the Netherlands, I see ChatGPT and Google AI Overviews as the two most relevant. Perplexity is growing, but remains smaller in the Dutch market for now.
Read also: Everything a B2B Marketer Should Know About GEO in 2026 · What is GEO?