Reputation, presence, relationships: the brands that thrive don’t wait for their audience to show up. They go out and find it.
The way people find information, discover brands and make purchase decisions is changing faster than you think, and many companies are failing to keep up.
Google’s AI Overviews, ChatGPT-generated results, Reddit threads climbing the SERPs: online visibility no longer comes (only) from having the right page in the right place. It comes from being a brand that intelligent systems recognize as authoritative, consistent and trustworthy.
In this scenario, building a clear brand identity is no longer a long-term strategic positioning exercise to put off until later. It’s a precondition for existing in AI results.
If a language model doesn’t know who you are, what you do and why you deserve to be cited, it simply doesn’t cite you. End of story.
Organic traffic is being redrawn. Traditional metrics tell less and less of the real story. And the companies that keep measuring success only in sessions and rankings will find themselves optimizing something that no longer exists.
Today, for SEO Confidential, we have the pleasure of interviewing Emina Demiri, Head of Digital Marketing at Vixen Digital. Emina is one of those professionals who doesn’t specialize in just one thing. SEO, LLMs, PPC, CRO, automation: her skills span the entire digital marketing ecosystem, with a data-driven perspective that always starts from business goals, not vanity metrics.
In this interview, Emina debunks some of the most widespread myths about SEO in the AI era, explains why traffic is the wrong metric to watch, and says things many people think but few have the courage to put in writing.
Recommended reading if you want to understand how to become the answer AI chooses rather than just one result among many.
The brands that win are the ones that build relationships
If AI Overviews are cutting clicks to websites so drastically, isn’t it finally time to state clearly that the old SEO model based on rankings and organic traffic is no longer enough to measure the real value of online visibility?
Yes, and to be honest, it was time to say it even before AI Overviews. I’ve been arguing for a while that traffic shouldn’t be the only benchmark for KPIs. We recently worked with a brand that saw a 26% drop in sessions while achieving a 52% increase in revenue from SEO. If we had reported only session data, the client would have panicked and we would have tried to fix a problem that wasn’t even a problem. Not to mention it would have been considered a failure!
The old model (things like ranking for keywords, generating clicks, measuring sessions…) was always meant to be part of the mix, not the only thing in the mix. The question to ask isn’t “how many people visited our site?”, but “did the right people find us, and did they do something useful when they did?”.
Rankings and traffic are activity metrics. Revenue and conversions are business metrics. We’ve confused the two for far too long, and AI Overviews have simply ripped off the band-aid.
Today, a piece of content can be found, read and even deemed useful by AI systems, only to be left out of the final answer shown to the user. Isn’t this the clearest sign that the real competitive landscape has shifted from simple page optimization to the overall strength of the brand, its reputation and its authority?
Absolutely, but it’s not really a new principle. It’s just that now it’s impossible to ignore. In the era of conversational search and large language models (LLMs), there’s practically no other way to rank than through your brand. LLMs don’t just look at your page, they look at the sum of everything that exists about your brand across the web. Things like your presence on third-party sites, mentions in communities, the consistency of your message, your topical authority… All of this helps determine whether you show up or not.
But to be honest, it’s been this way for a while in “traditional search” too. Yes, there are some very niche industries that can still “optimize their position in search results”. But even before LLMs, that group was shrinking. With BERT in 2019, Google told us loud and clear that it had moved from keyword matching to language understanding (context, intent, nuance…).
Then came the “helpful content” updates, E-E-A-T and the relentless series of core updates that kept penalizing thin, ranking-only content while rewarding genuine topical authority and real-world expertise. The direction of travel has been the same for years now. AI Overviews haven’t changed the destination, they’ve simply accelerated the journey dramatically and made it impossible for anyone to pretend they didn’t see the signals.
And now, on top of all that, you can even launch a website in a day with no prior development experience using large language models (LLMs) like Claude; too many people are churning out low-quality AI content and the market is more and more saturated.
So yes, you need a brand, and you needed one even before GenAI came into play. If you want to succeed in search, you should do what you were supposed to do from the start: understand your audience, your business, how you make money… Build the brand with a clear identity aligned with that audience, demonstrate deep authority in your niche, and make sure you’re consistent and present across multiple platforms. In short, get the fundamentals right!
This is actually all good news if you’re willing to do the harder, long-term work. It’s bad news if you’ve treated SEO as a purely “can you optimize this content for search engines” exercise.
In a scenario where search generates fewer clicks and brings less traffic to the site, which metrics are actually useful for understanding whether SEO activities are producing concrete value in terms of conversions, engagement and revenue?
Well, certainly not traffic, that’s for sure!
Before we even talk about metrics, we need to be honest about something more fundamental: the ability to measure anything meaningful depends entirely on your level of data maturity.
If your consent management is broken, tracking fires before consent is given, Google Analytics 4 was set up with a simple copy-paste and no custom events, and reporting is a huge manually compiled Excel file… you don’t have a metrics problem, you have a foundations problem.
Data maturity isn’t a luxury or a nice-to-have. It’s the infrastructure that makes everything else possible. That means a proper consent setup with Consent Mode, a measurement framework that ties real business goals to specific KPIs, a tracking plan built on those goals, and dashboards that connect directly to the data, without requiring reports to be built manually every month.
When you assess a client’s data maturity and the score is, say, 1.8 out of 5, that immediately explains why they feel unable to prove the value of SEO. It’s not that the value doesn’t exist, it’s that they can’t see it.
The measurement framework is the starting point for climbing that maturity ladder. That means running a workshop with the marketing, sales and customer care teams, the agency and anyone who works with data, to define what success really means at the business level. You start from business goals, move to marketing goals, identify the main KPIs, then the secondary metrics and segments. Only then do you build the tracking plan and the reporting layer. Doing it the other way around leads to dashboards full of data that don’t answer any real question.
Some metrics that usually fit this approach are: the conversion rate from organic traffic, revenue generated by organic, engagement rate, micro-conversions (such as email sign-ups, downloads, time on key content) and direct brand searches. That last one is crucial: when AI shows the brand without generating clicks, the result often surfaces later as a branded search. Monitoring its growth gives you proof that awareness is rising, even when traffic isn’t.
Then there’s another layer: behavioral data. Tools like Hotjar (now Contentsquare) or Microsoft Clarity, alongside GA4, let you really observe user journeys. Analyzing heatmaps and behavior is essential. It’s surprising how many SEO professionals simply report acquisition data without looking at what happens after the click. That’s exactly where the most important part of the story lies, and it’s exactly where SEO and CRO need to come together. With less traffic available, every visit becomes more valuable. Every session counts more than before. The job is to make sure the site deserves it.
And, please, stop trying to report on AI rankings. There’s no such thing in AI-powered search. It’s just theater.
In terms of time and budget, which activities should an SEO team focus on today, and which ones are overrated compared to the results they actually deliver?
Focus on: deep audience research. Truly understanding users’ real motivations, not just their search queries. Building topical authority in specific niches instead of chasing high-volume keywords. Betting on content distribution and repurposing: every major asset should be reworked into at least five different formats. Updating content that already performs instead of endlessly creating new pages.
For eCommerce, paying attention to the product feed is essential. Too many SEOs working in eCommerce don’t engage with the product team and often don’t even collaborate with PPC to optimize the feeds. In the past that could just about work, partly thanks to organic shopping, which many ignored for far too long. Today, with the rise of agentic eCommerce, it’s no longer an optional skill. Optimizing the product feed and aligning it as closely as possible with schema markup is becoming increasingly important.
And then CRO: it’s an integration that has to happen. SEO and conversion rate optimization must work together. Agents are far less tolerant than humans when it comes to friction in navigation journeys.
Overrated: the obsession with word count. Chasing generic high-volume keywords. Link building strategies with no real understanding of the audience. And today, as already mentioned, the biggest waste of time is teams spending hours trying to “track AI rankings” or reverse engineer LLMs, devoting far too much time to trying to “chunk” content artificially. You cannot reverse engineer a neural network. There’s little point in continuing to try: that time is far better invested in truly understanding your audience.
The underlying rule is simple: instead of obsessing over “chunking”, talk to your customer service team or your salespeople. Those conversations are worth more than any simulation.
When a company needs to rethink its website, content and digital strategy, what are the most effective steps to truly understand what its audience is looking for, how it gathers information and what drives it to convert?
You don’t start from an Excel sheet full of keywords. You start from the audience. It may sound obvious, yet many strategies are still built top-down, based on the brand’s assumptions rather than users’ actual behavior.
At Vixen Digital we take a three-layer approach. The first is going where the audience already is: Reddit, communities, forums, social networks. There, you analyze conversations to understand the language people use, the problems they express, what builds trust or distrust. This work can be done systematically using tools like the Reddit APIs together with topic modelling techniques.
The second layer is talking directly with the sales and customer service teams, or better still with customers themselves. It’s essential to understand which objections come up, which questions get asked and which terms resonate most. This is extremely valuable first-party data, and it’s often overlooked.
The third layer is using language models to synthesize and organize data at scale: intent classification, sentiment analysis, pattern recognition. This is where LLMs deliver the most value, letting you process large amounts of information. It remains essential, though, to understand how the different models work, in which contexts they’re effective and where they show their limits.
Not everything is suited to LLMs. A useful reference is Lazarina Stoyanova‘s framework, which helps you assess the problem before automatically reaching for a model as the solution.
Only after deeply understanding the audience and the business can you build the site and content architecture. It has to reflect how the audience thinks and searches for information, not how the company describes its products internally. There’s an effective analogy here: many hygiene programs in developing countries failed for years because they didn’t start from local communities. The same principle applies in marketing: if you don’t start from the audience, you risk building something nobody asked for.
And of course, you need to set up a measurement framework to track and analyze user behavior on the site. It’s a point that has already come up, but it remains central.
With all the attention today focused on AI, fan-out, chunking and other technical concepts, how can we avoid the risk of creating content for algorithms instead of for the people who need to read it, trust the brand and take action?
That’s exactly what worries me most.
The industry has been through phases like this before. We had keyword density, domain authority (backlinks), and now there’s the chunking obsession. Every time a new technical concept emerges, people start creating content for that concept instead of for the human being who has to read it, trust it and act. I mean, in the case of chunking it doesn’t even make sense! (A good resource on the myths around new tools is the one by Dawn Anderson).
The curiosity to cultivate is this: “what does this person really need in order to understand or do something?”, not “how do I structure this for fan-out?”.
Don’t get me wrong: technical knowledge is extremely useful, and it’s also fun. I openly call myself a technical nerd! Understanding how LLMs process context, how embedding space works, how entities connect… all of this matters. But it has to serve the content strategy, not drive it.
If you find yourself writing for “chunks” instead of for understanding, you’ve gone in the wrong direction!
It all comes back to what we said earlier about the audience. If you start there, you automatically end up serving people rather than algorithms. And when that part is done well, you very often end up serving both!
Recently we’ve seen an extreme case: webmasters producing .md versions of their sites in the hope of pleasing the machines. It’s madness!
Jono Alderson explains it well: a page is more than a container of words. Flattening content into a machine-readable format doesn’t give LLMs more signal, it strips away the editorial hierarchy and intent that meaning depends on. The answer isn’t a “shadow” version of your site, it’s a better version of your site, full stop.
Another very effective way to think about the needs of machines and people is Ramon Eijkemans’ “utility writing” approach. The core idea is fairly simple: instead of “bolting” structured data onto content as an extra technical layer, you write language that is inherently machine-processable from the start. Things like named entities, explicit relationships, preserved conditions, self-contained sentences… you keep the structure machines need directly in the language itself. What’s most striking is that this approach completely removes the false tension between writing for humans and writing for algorithms.
If every sentence stands on its own, if you make relationships explicit instead of just listing entities, if you include at least one statement an LLM could quote directly, you get content that is clearer, richer and more useful for everyone. And crucially, as Ramon Eijkemans points out, this is not some new “GEO” trick!
Google has been evaluating content at the passage level ever since passage ranking was introduced in 2021. The information retrieval infrastructure is the same. You’re not optimizing for different systems, you’re optimizing for a single evolving system that rewards well-structured language.
So, if you start from the audience, build a clear and consistent brand, write with real utility, create better websites (not AI slop!) and make sure every sentence does its job, you end up serving people and machines at the same time.
If Google, ChatGPT and even Reddit are turning the decision and purchase phase into a process that happens entirely inside their platforms, aren’t we witnessing a progressive expropriation of the direct relationship between brands and customers?
Yes and no. These platforms are evolving into spaces where research, influence and purchase decisions increasingly happen in the same place. It’s a significant structural shift, and brands need to respond to it strategically, not reactively.
The answer isn’t to panic, but to diversify and build depth. It’s really nothing new. The brands that have always worked are the ones that show up consistently across multiple touchpoints, not the ones that put all their eggs in one basket. It’s the basket that’s changing shape, once again.
What you can do is make the brand clear, present and trustworthy across the whole ecosystem (owned, earned and paid channels). That way, when a user encounters the brand through an AI summary or a Reddit discussion, a certain level of recognition and trust already exists. The brands that manage to endure are the ones that build genuine relationships with their audience and optimize their presence across multiple platforms. That also means going beyond the obvious channels and prioritizing first-party data and communities.
The platforms aren’t taking customers away. They’re simply changing where the conversation starts. The job is to make sure the brand is part of that conversation, and that it doesn’t end there.
Advertising inside ChatGPT offers immediate visibility, but it ends the moment the investment stops. In which cases can it be a strategic lever, and in which does it risk turning into an economic dependency with no lasting value?
It’s hard to say for sure, since we’re still in the early days. But it’s likely that many of the dynamics we already know from more established paid channels will apply here too.
Paid visibility, in any channel (not just ChatGPT), is a lever that should support a long-term strategy, not replace it. If you use ChatGPT ads to amplify a brand that already has strong organic authority and a clear strategy on its owned channels, it can be very effective. Especially in high-intent, time-sensitive moments where visibility at the decision point really makes the difference.
It becomes a trap when it replaces the harder work, which is building a brand. “Appearing in AI answers by paying for placement”, without a long-term vision and without a solid brand, ends up being a cost line, not a strategy. The moment you stop investing, you disappear. That’s the economics of dependency, and too many companies have already learned it the hard way with Google Ads. Without an organic foundation that compounds value over time, the numbers rarely add up at scale.
Paid visibility in AI should be treated like any other paid channel: you have to ask whether it’s supporting or replacing organic value. If it’s the latter, you have a problem.
This race among Big Tech players to become the place where purchase decisions are made risks wiping out years of investment by companies and publishers in SEO, content and owned websites. Are we facing a new form of intermediation that concentrates even more power in the hands of a few players?
What we’re seeing is the formation of an additional layer between brands and customers along the entire journey: from discovery, to research, to evaluation, all the way to purchase. These platforms shape which information gets shown, in what format and with what context. That’s not necessarily a bad thing: it’s how information systems naturally work at scale. But it does mean brands need to think carefully about where they place their dependencies.
The SEO industry has been through similar dynamics before. Just look at organic reach on social: once enormous, now almost nonexistent. Does that mean you should stop doing organic social? Or is it really surprising that profit-driven platforms push toward paid? The answer is obvious.
What’s changing today is the speed and scale of the phenomenon. AI-driven search isn’t a simple channel update, but a deeper shift in how people access information. The intermediary is increasingly becoming the interface itself.
The strategic answer, though, remains the same, only more urgent: build direct relationships with your audience that don’t depend on a single platform. Owned channels, email, communities, first-party data… these have always been fundamental. Now they become an even more critical strategic asset.
The brands that navigate this best won’t be the ones fighting the platforms, but the ones using them strategically while building something that can’t be replicated: a direct, trust-based relationship with their audience.
Reddit was born as a space for authentic discussion among users, but today its conversations are turning into commercial suggestions and potential sales channels. Isn’t there a risk that community trust gets monetized to the point of completely compromising its credibility?
It’s a real tension, and that’s coming from someone who uses Reddit data extensively for audience research!
The authenticity of those conversations is the whole point. Reddit is valuable precisely because, historically, it wasn’t built for commercial intent. People say what they actually think. They describe real problems in real language. Of course, it also depends on the specific subreddit and on the work of the moderators… but that’s another story!
As the advertising side expands and more and more companies inject sales-oriented content, it’s inevitable that organic conversations get “contaminated” by commercial intent. This can affect user trust, but it doesn’t automatically mean a mass exodus from the platform. Just look at how Facebook evolved.
For now, the authenticity is still there (though it depends a lot on the subreddit). The advice for brands is to use it as a listening tool and treat it with respect, like any community, avoiding building an artificial presence. Communities sense it immediately. Credibility comes from genuine participation. The moment you start “optimizing Reddit” the way you optimized your website, you destroy exactly what made it useful, (as Erin Simmons put it brilliantly in her excellent talk Trust > traffic: how community fuels organic growth).
Write for people, build for machines, invest in the brand
Brand, authority, genuine relationships, first-party data. These aren’t new concepts.
They are exactly what should have been guiding digital strategy even before AI became the main interface between people and information. The difference is that today there’s no room left to pretend you didn’t see the signals.
Know your audience.
Go where it is.
Listen to how it talks, what it looks for, what holds it back.
Only from there can you build content that truly holds up: for the people who need to trust you and for the machines that decide whether to cite you. It’s not a tension to manage. It’s a natural synthesis that comes when you stop thinking about algorithms and start thinking about people.
Those who have built something solid, a recognizable brand, a consistent presence across multiple platforms, an audience that searches for them by name, are already in the right position. Those who optimized pages without building an identity now find themselves having to start over.
The right time was a few years ago. The second-best time is now.
A heartfelt thank you to Emina for her dense, precise answers.
And as for us? Well, see you next week with another interview you won’t want to miss, right here as always, on SEO Confidential.
