SEO Confidential – Our exclusive interview with Simon Schnieders: in a world without clicks, the brand is the result

With generative AI synthesising answers and shaping users’ memory, the challenge is no longer just being found, but being interpreted correctly by the model

I created SEO Confidential to have a space for open, direct discussion, with no filters and no convenient stances. A place to genuinely debate what is changing in SEO and, more broadly, in the way brands, content and reputation are interpreted by search systems and artificial intelligence models.

If you follow me, you know very well that I sometimes disagree with the guests I interview, but I believe this diversity is an asset that only does the debate good.

It’s no secret, for example, that I look at the definition of GEO with a certain wariness and still consider it, in essence, a natural evolution of SEO rather than a standalone discipline. But, at the same time, I believe it’s more productive for everyone to stop hiding behind labels, which are only good for wasting time in ideological battles.

That’s another reason why I’m particularly happy to host Simon Schnieders, founder of Blue Array, a UK SEO (and GEO) agency, one of the industry’s most authoritative voices and a staunch defender of GEO.

Over the course of the interview we will tackle topics that closely affect professionals like you, brands and publishers: the progressive loss of meaning of rankings, visibility without clicks, the role of sentiment and bias in generative models, the survival of brands in an ecosystem dominated by synthesis (remember AI Overviews?) and much more.

But above all we will talk about corporate reputation and how AI can rewrite it, even incorrectly. A risk Simon experienced first-hand, when Blue Array was hit by a scam capable of distorting the public perception of the brand.

His response to that attack was, in my opinion, excellent, and I seriously think you should know about it.

“In a world without clicks, the brand is the result. If the model trusts you, it mentions you. If it mentions you, the user searches for you directly,” says Simon

Hi Simon, I recently read a post of yours on LinkedIn in which you write that “you can’t rank in a probabilistic system” and that the tools promising to measure a brand’s “visibility” in LLM answers produce noise, not data.

In your view, are we facing a new bubble of vanity metrics, or is there still a legitimate space to monitor, in some form, brands’ presence in LLM answers?

We are definitely facing a snake oil bubble (the expression comes from the supposed miracle “snake oil” remedies sold in the nineteenth and early twentieth centuries, especially in the United States, marketed as universal cures but with no real efficacy, Ed.).

The fundamental mistake many new “AI visibility” tools make is treating a Large Language Model like a deterministic database. In classic SEO, if you rank first for a keyword, you generally rank first for everyone in that region.

LLMs, on the other hand, are probabilistic. Ask ChatGPT the same question five times and you might get three different answer variants. So a tool that promises rank tracking based on static prompts is just selling you noise.

That said, presence still needs to be monitored, but the reference metrics change. We need to move away from “rankings” and towards Share of Voice (SoV) and Sentiment. The question is not “where do I rank?” but “across 100,000 query iterations within this ‘category’, in what percentage of cases is my brand mentioned or cited?” and “is that mention or citation positive or negative?”. That is data; the rest is vanity.

GEO (Generative Engine Optimization, optimising content for AI-powered answer engines) is often mocked as “a made-up name”, but you think it is a natural evolution of SEO, right? Why, in your view, would it be a mistake to deny that GEO exists?

To me, the mistake many SEO colleagues make is thinking of GEO as a “new channel” rather than the natural evolution of the channel they already work in. GEO is not an invented buzzword; it is simply the shift from optimising for retrieval (finding a document) to optimising for synthesis (generating an answer).

The SEOs who deny this are buying into Google’s propaganda that “it’s just SEO”. I have seen hard data suggesting that around 50% of AI answers come from the base model rather than from content retrieved in real time.

This means that a significant share of visibility and informational framing carries the biases and preconceptions baked into the model itself. Acknowledging this doesn’t mean abandoning SEO, but recognising that the context it operates in is changing.

You have suggested that classic SERPs could disappear by 2027 and that in the future generative models will entirely decide what to show, often without clicks to websites. In your view, which categories of brands risk disappearing first from the models’ “memory”? And which are better positioned to survive in a world dominated by brand visibility and sentiment?

It’s hard to imagine a future for many “intermediaries”. When the business model rests solely on information arbitrage (that is, on reorganising content produced by others without any testing, analysis or direct production, Ed.), as happens with affiliate sites listing the “best toaster ovens” without ever having tried one, the risk of being pushed out of the market becomes real.

LLMs can in fact do that synthesis work instantly and, in most cases, more effectively.

The survivors will be Source of Truth brands. These are brands with “Real World” signals: physical footprints, proprietary data, active communities and authentic reviews. At Blue Array, being a B-Corp helps in this respect because it is a verified external signal of legitimacy. In a world without clicks, the brand is the result. If the model trusts you, it mentions you. If it mentions you, the user searches for you directly.

You maintain that GEO exists and is already part of the present, but you also admit that visibility in generative models is almost impossible to measure reliably, because LLM answers change constantly and are inherently probabilistic.

In this context, how do you prove results to clients? Which metrics, signals or forms of evidence do you find genuinely useful to show that GEO work is having a real impact, even without a “ranking” to monitor?

This is the hardest conversation we are having right now. We have to “re-educate” clients to look at different signals. We focus on correlative visibility.

We can’t track a static ranking, but we can track:

1. Brand mentions in AI experiences: we use manual sampling and emerging tools to see whether the brand appears in the “snapshot”.
2. Traffic quality vs. quantity: we may see a drop in organic traffic volume but a rise in conversion rate. Why? Because the “curious” got their answer from the AI, so the people who do click are ready to buy.
3. Third-party citations: we show clients that we have earned coverage on the sources that teach the AI. If we influence the source, we influence the output.

You wrote that, when a user asks “what’s the best X”, LLMs don’t look at the word “best” superficially — they don’t think about rankings or the best keyword, but activate a set of criteria such as quality, price, durability, reputation, reviews. What, in your view, is the most common mistake marketers make when thinking about the concept of “best”?

They think “best” is a keyword. They try to optimise a page by stuffing “Best SEO agency” into the copy.

For an LLM, ‘best’ is a set of attributes or vectors. When a user asks for “the best running shoes”, the LLM looks for specific vector associations: durability scores, opinions in Reddit threads, return policies and price consistency.

The mistake marketers make is saying “we’re the best” instead of demonstrating the attributes the model associates with quality. You need to decode the criteria the LLM uses to define “best” in your vertical and optimise for those criteria, not for the adjective itself.

Google AI Mode uses query fan-out to turn a single question into many sub-questions, so as to better understand what the user really wants. From your point of view, what does this change in logic mean for brands? Who risks losing visibility, and who stands to gain it, once Google no longer interprets queries literally?

“Query fan-out” is fascinating because it kills the long-tail strategy of creating a page for every specific question. Google now breaks a complex question into sub-parts, retrieves information from different sources and stitches it together.

Brands relying on “thin” content that answers simple questions risk losing visibility because Google no longer needs to send a user to their page to answer that specific nuance. The winners will be the brands that provide comprehensive coverage. You need to be the authority on the entire topic so that, when Google breaks that query apart, your entity is the answer to 3 or 4 of those sub-questions.

During the Simply Business event in London, in the panel on SEO and AI that you moderated, Jonathon Heard, a Google UK manager, said that the introduction of AI is driving users to click more and dig deeper into their searches.

It’s a statement that clashes with what many publishers and businesses are seeing in real data, namely sharp drops in organic traffic. How do you interpret this gap? Does it strike you as a show of optimism, or do you genuinely believe that the use of Gemini 3 can increase clicks to external sources?

I have a lot of respect for Jonathon, but that statement is, let’s say, “corporate optimism”. The data we see across our client base (and in the industry at large) doesn’t support the idea that AI generates more clicks for everyone. If anything, it generates zero-click satisfaction for simple queries.

That said, there is a grain of truth: for complex, in-depth topics, users may click through to more qualified sources. But for the vast majority of publishers relying on shallow traffic? No. That traffic is being cannibalised. We have to be realistic: Google’s goal is to satisfy the user on Google.

You asked Google whether it will ever be possible to get separate data in Search Console for AI Mode and AI Overviews. The answer was a generic “we’re thinking about it”. Why this stubborn lack of transparency? Is Big G afraid those numbers might show a negative impact for publishers and businesses, or is it the very structure of the new Search that makes it hard to obtain separate, genuinely reliable metrics?

It’s probably a mix, but I lean towards strategic obfuscation. Technically, separating the data is possible — they know exactly where the click comes from. But if they explicitly showed publishers: “Look, 40% of your impressions are now strictly within AI Overviews with a 0.1% CTR”, it would cause a revolt.

It would devalue their ecosystem. By blurring the boundaries, they buy time to normalise the behaviour. It’s frustrating for those of us in data-driven marketing, but from the standpoint of Google’s business strategy, it makes perfect sense.

Now, Simon, let’s move on to a topic very close to my heart. With the rise of LLMs there is a risk of a sort of “parallel brand” taking shape, made of hallucinations, wrong information or simple misunderstandings that can take root in public perception. How real is this danger and, in your view, what strategies should a company adopt to actively influence generative models and make sure the brand narrative is the real one, not the one invented by the AI?

The danger is incredibly real. If an LLM hallucinates that your product is discontinued or unsafe, that becomes the “truth” for thousands of users.

The strategy is “Inception”. You don’t correct an LLM after it has answered, because you can’t edit it directly. You act beforehand, influencing the information the model learns. How? By flooding the training corpus with structured, verifiable data.

1. Schema Markup: be obsessive about Organization and Product schema. Speak the machine’s language.
2. Wiki strategy: make sure your presence on sites like Wikipedia, Wikidata and Crunchbase is accurate. LLMs trust these sources disproportionately.
3. Digital PR: you need your brand narrative present on high-authority third-party sites. If the New York Times says you are X, the LLM believes you are X.

Blue Array, your consultancy, was the victim of a scam that generated negative conversations online and risked compromising the brand’s reputation in the eyes of LLMs as well. What steps did you take to counter this distortion, rebuild trust and get the models to tell the correct version of your identity?

This was a critical test of our GEO theories. We were the victims of a “scam” in which bad actors used our brand on WhatsApp to defraud people. Negative comments started spreading on Reddit, with people asking “Is Blue Array a scam?”.

We realised that if LLMs picked up those Reddit threads, our brand would be associated with the word “scam”. We took a multi-pronged approach:

1. Direct intervention: I personally went into those Reddit threads to clarify the situation, confirming that it was a scam and that we were its victims. This added an “authoritative” text to the thread.
2. Content creation: we published a clear, detailed page on our website about the scam.
3. The result: now, if you ask ChatGPT or Gemini about the “Blue Array scam”, they won’t tell you that “Blue Array is a scam”. They say instead: “Blue Array was targeted by scammers and warned users about it”.

We successfully taught the “AI brain” to distinguish between the criminal and the victim by giving it the right context in the places where it looks for the truth (like Reddit and our website). That, to me, is GEO in action.

Below you will find the video in which I comment on Simon Schnieders’ interview and tell you why it contains vital information for your business:

When AI rewrites your brand, strategy becomes a responsibility

This conversation, like all the SEO Confidential interviews for that matter, is about you, your brand, the decisions you are making today and the ones you keep postponing because “it’s not clear yet”.

The point is that it never fully will be.

Generative models are already filtering, synthesising and rewriting reality, and they do so whether or not you believe in the new labels.

You can keep chasing increasingly unstable rankings, or you can start asking yourself what image of your brand is emerging when nobody clicks any more.

You can rely on reassuring but empty metrics, or accept that visibility, trust and reputation have become variables that are harder to measure, yet far more decisive.

I think you should keep in mind a concept that emerged in this interview: you don’t control the model, but you can influence what it learns. And that requires intentionality, presence and consistency. In the content you produce, in the sources you oversee, in the conversations you choose to be part of.

Corporate reputation is no longer just the result of what you communicate, but of what artificial intelligence systems learn, synthesise and hand back to the public. It is a silent but profound shift.

The hallucinations, biases and simplifications of generative models can turn a marginal piece of information into a perceived truth, amplified at scale.

It doesn’t take a glaring error: an incomplete context or a lopsided source is enough.

In this scenario, reputation is not defended by reacting after the fact, but by building a solid, consistent and verifiable information ecosystem beforehand. Those who underestimate this shift risk losing sovereignty over their own story just when they believe they have it more under control than ever.

If you work in digital, you can’t afford to stay neutral. You have to choose where to position yourself, even when positioning is no longer a ranking.

This is not the end of SEO, nor the beginning of a new acronym to adopt. It is a shift in responsibility.

And ignoring it, today, is the riskiest choice of all.

Well, I thank Simon for his precious insights and his availability and, as for us, see you next week with a new interview you won’t want to miss!

#avantitutta!