“The web has never been a meritocracy. It has always been survival of the fittest. Whoever adapts fastest wins”
Our journey continues with SEO Confidential, the column that gives a voice to the protagonists of change: international experts who analyse, discuss and sometimes clash over the transformations of online search.
At a time when attention is shifting from search engines to AI- and LLM-powered answer engines, this series of interviews gathers concrete, technical and strategic points of view for business owners who want to keep selling and converting in a digital ecosystem that changes every day.
In this episode, our guest is Duane Forrester, one of the world’s most authoritative voices in search and digital visibility. A former product manager at Microsoft Bing, where he launched Bing Webmaster Tools and contributed to the birth of Schema.org, he has spent more than twenty years studying how people – and now machines – find, interpret and judge online content.
Today he guides companies and startups through the era of GenAI discovery, helping them become discoverable, trustworthy and retrievable within artificial intelligence systems.
In this interview, Duane explains how SEO is changing in the age of answers generated by language models, why data matters less than it used to, and how a brand’s credibility is becoming a fragile balance between algorithms, licensing deals and human perceptions.
We’ll talk about Machine-Validated Authority, bias in models and a web that perhaps was never truly meritocratic, and much more…
A conversation I believe you’ll find very useful, one that invites you to look beyond the click and understand how artificial intelligence is rewriting the very concept of online visibility. Enjoy the read.
“Universal verifiers are coming soon: they will check facts before generating an answer. Hallucinations will decrease and trust in the platforms will grow”
We often hear about “data-driven marketing”, yet with AI assistants no one really knows what gets seen, by whom, or how much it actually matters. Isn’t that a paradox? Are we really measuring anything, or are we just pretending to understand numbers that no longer make any sense?
There’s a lot to unpack in that question, so let’s start at the beginning. The platforms know perfectly well what gets displayed. They have the data on what is shown, how often, in what form and to whom. They know when a piece of content is cited, mentioned, paraphrased or left out of an answer entirely. And when users are logged in, they know exactly who is seeing it. Even when they aren’t, they can still analyse the behaviour of user cohorts with precision.
So yes, the platforms know, and that data matters to them. It drives decisions worth millions, from product features to API access to what gets prioritised in results. The paradox isn’t that we have no data. It’s that they have more of it, and we have less.
For the rest of us, it’s complicated. SEOs live in an approximation of the truth. Some understand what is really happening. Most keep guessing. But that’s how SEO was born, right? In the early days, we were all guessing. Right now, we’re back in the same position, staring at a new system and trying to reverse-engineer its rules.
Those who learn how AI systems actually make decisions, who dig into how models evaluate, score and prioritise content, will succeed. Everyone else will fall behind, still staring at dashboards built for a world that is ceasing to exist
Google has started blending AI Mode data into Search Console, but without letting you separate it from traditional search data. Do you think this is just a temporary technical limitation, or a deliberate choice to avoid showing how much traffic AI is already taking away from classic results?
In my view, it’s a decision, not a limitation. From a business standpoint, there are plenty of good reasons not to disclose that data. If Google let everyone see exactly what share of clicks or impressions is being intercepted by AI answers, it would hand competitors a blueprint. Any company could reverse-engineer it and accelerate the development of its own version.
And the cost wouldn’t be trivial. When you factor in the product management, engineering, design, testing and QA work required to build and maintain that kind of data segmentation, you’re talking hundreds of thousands, maybe millions, invested in creating that layer. Sharing it openly would be like giving away the recipe.
Would it help SEOs and marketers? Barely. A bit more information would help, sure, but it wouldn’t really change strategies in any substantial way. That’s why I believe it’s a well-considered business decision, not a technical limit.
Everyone pushes us to “write for users”, yet AI assistants reward short, structured, standardised text. Aren’t we at risk of building a web designed to satisfy language models rather than people?
Let’s look at it from another angle. Do we really believe the web should exist only for human beings? The entire foundation of the web was about sharing information between sources, not necessarily people. Machines are now part of that ecosystem.
If my goal is to earn money and pay my employees, I care about the outcome, not the species of my visitors. Whether the visitor is human or machine, if it consumes my data, cites my work or buys my product, that’s a win.
It’s not about whether we like it. It’s happening. And it isn’t going back. We can hate the idea, we can argue it’s “soulless” or “wrong”, but billions of consumers use AI every day. When demand is that high, resistance is futile. (And yes, I’m using a Star Trek reference!) So, do we want to get off the train or learn? Adapt half-heartedly, or use the chaos to our advantage?
“Artificial intelligence doesn’t replace SEO, it complements it“. And yet, if answers appear before the click even happens, isn’t that a sign that SEO has already lost its most authentic function: generating real traffic and visibility?
I think of this transition like moving from high school to university. You can’t do one without completing the other.
Is SEO losing value? It depends on what you mean by value.
If you believe SEO’s value lies in the size of the industry (the jobs, the conferences, the tools, the agencies), then yes, this evolution challenges that model.
But if you believe SEO’s value lies in improving access to information, making content clearer and ensuring trustworthy experiences, then no, it isn’t losing value.
AI systems reward structure, clarity and factual alignment. That is exactly the kind of work good SEOs have always done.
The difference is that the “audience” now includes machines. If you’re optimising for retrieval and trustworthiness rather than rankings and clicks, you’re still doing SEO. You’re just doing it for a new, deeper and more interconnected layer of search.
AI assistants decide who gets cited, but the criteria remain unclear and potentially shaped by licensing deals with publishers and media outlets. How can we talk about meritocracy if visibility can depend on who signs a contract with OpenAI or Microsoft?
Broadly speaking, those licensing deals do carry some influence. But publishers aren’t necessarily the most knowledgeable experts on most topics. When a query calls for real expertise, the playing field levels out again.
We’re used to thinking about search in linear terms: “if I do X, I get Y”. AI models operate in a vector space, where relationships are multidimensional and constantly evolving. Trying to apply the old SEO logic to that world doesn’t work.
And yes, the systems are opaque. But so was Google in 1998. We learned, we adapted, and we built entire careers out of that opacity. There’s no difference.
Besides, let’s be honest… the web has never been a meritocracy. It has always been survival of the fittest. Whoever adapts fastest wins. And if you think about it, even the concept of meritocracy implies competition. One winner, many losers. The web has always played by those rules.
You talk about “Machine-Validated Authority” as a new indicator of trust. But if no one can truly verify it, isn’t there a risk it becomes the new illusion of ranking? A concept designed to lend credibility to a system that, in practice, no longer lets us understand how it evaluates content?
No one outside the platforms will ever see how AI systems assign trust. That doesn’t mean the signals don’t exist. Of course they do. Models constantly evaluate credibility, consistency and accuracy at a level of granularity we’ve never had access to.
Our job is to present content in a way that conveys every possible signal of trustworthiness. The more we help the machine understand our expertise, the better our chances of being chosen.
The next phase, and it’s coming soon, will be these companies rolling out universal verifiers. These are systems that check the facts before delivering an answer. That will drastically reduce hallucinations and increase trust in the platforms.
When that happens, low-quality or misleading content will be filtered far more aggressively. The signal-to-noise ratio will improve, trust will grow, and the content that survives will earn machine validation based on merit, not manipulation. Well, something like merit, given that we’ve already called that concept into question…
Consumers will reward the platforms that deliver the most accurate results. Once that loop closes, debating whether the system legitimises itself will become academic. The market will decide what works.
Today, many practitioners accept every statement coming from Google or OpenAI without question. Isn’t that a sign that SEO has stopped being an experimental, critical discipline and turned into an act of faith towards whoever controls the platform?
Honestly, I’m not quite sure when all this started. When I worked at Microsoft Bing, I questioned everything. When we had to decide what to make public, we brought together engineers, lawyers, PR people, product managers and everyone else. If the message risked being too shallow or misleading, we said nothing. That way, when we did speak, people knew it mattered.
Today I see a lot of repetition. Someone from Google says something and it gets amplified immediately, often without context. Repeating isn’t the problem; repeating and believing without verifying is.
It’s like a beer company telling you its beer is the most refreshing in the world. If you’ve never tasted another one, how would you know?
There used to be more pushback. SEOs held Google to account, loudly, at conferences and online. That energy has faded. Maybe it’s maturity, or maybe it’s fatigue.
But if you still believe your future lies entirely in “traditional SEO” and you ignore how AI is redefining the interpretation of data, then I don’t know what to tell you except, perhaps: goodbye? Technology drives change, and standing still guarantees falling behind.
You argue that in 2026, AI assistants will become the first point of contact with information. If consumer and business decisions are made before people ever reach websites, isn’t there a risk that the battle for visibility moves off the web for good?
No, not at all. Shifts like this don’t happen overnight. Human behaviour evolves slowly, even when technology moves fast. After all, it took Google nine years to reach a 50% market share. AI adoption is moving faster, but it’s still far from instantaneous.
What will really change is how visibility manifests itself.
Websites won’t disappear, they’ll transform. Data still needs a home, and content must remain findable, indexable and verifiable.
AI assistants don’t create knowledge out of thin air: they feed on a well-structured, well-optimised web.
Some companies may abandon the traditional website model, but their online presence will live on in other forms. The real challenge will be writing and optimising for two distinct audiences: human beings and the AI crawlers that analyse, interpret and validate information.
If AI models can’t tell the difference between truth and repetition, doesn’t that mean a brand’s reputation can be destroyed simply by whoever manages to pollute the data enough?
There’s a kernel of truth in that. The problem exists, but a great deal of work is going into solving it. All the major companies are building systems that verify AI outputs using other AIs and, ironic as it may sound, it’s proving effective.
When you narrow a model’s data to verified sources, hallucinations drop dramatically. The trade-off is flexibility, but accuracy goes up. The “universal verifier” concept is designed precisely to handle this.
Sure, right now a wave of bad data can hurt a brand. But that window is closing. We’ve been through the same thing with negative SEO and toxic link schemes. There was chaos, then the problem got solved. The same will happen here.
According to what you write, a few hundred pieces of fake content are enough to sway an AI assistant’s judgement. So who guarantees that a brand’s “credibility” doesn’t become just a matter of volume and manipulation?
Guarantees? None. There never were any. But the responsibility is shared.
It all starts with the brand: leadership, marketing, communications and even product teams all contribute to how a brand builds trust. Then there are fans and advocates, who can amplify its reputation, sometimes strengthening it, sometimes weakening it.
Platforms play a key role too. Google, Bing, Yelp and AI assistants must provide systems capable of verifying authenticity and preventing manipulation. So-called “universal verifiers” could reinforce that balance, but we’re still a long way from a reliable model.
Then there are journalists, influencers and bloggers: voices the algorithms read as signals of authority. All of this shows that a brand’s credibility is a collective effort, not an isolated task.
We’ve been dancing this dance for decades. The tools and platforms change, but the fundamentals stay the same. Those who truly understand reputation management will keep thriving.
If every search engine and AI model is inevitably shaped by bias, does it still make sense to talk about “neutral results”? In practice, who decides how a brand is perceived: the algorithm, or those best equipped to influence it?
I’ve never believed in neutral results. Algorithms have always reflected the biases and intentions of their creators. AI models are no different, they just do it faster and at scale.
The algorithm builds the environment, but it’s human beings who shape the signals that feed it. Every click, share or interaction helps reinforce the model’s view of the world.
Who really determines how a brand is perceived? Both.
The algorithm sets the rules of the game, but human behaviour writes the match. Those who can read and steer that relationship will be able to shape perception. The rest will stay a step behind, forced to chase.
Bias, perception and power: who really controls online reputation?
The interview with Duane Forrester opens a lucid, almost unsettling perspective on what lies ahead of you.
A future in which purchase decisions happen before the click, answer engines drive discovery and artificial intelligence redraws the very concept of conversion.
The certainties of the past – keywords, rankings, CTR – are giving way to a new ecosystem in which visibility is no longer measured only in clicks, but in trust signals recognised by machines that learn, synthesise and judge.
You should know that today a brand’s credibility is the product of a ménage à trois: what the company communicates, how the public reacts and how the algorithms interpret both.
Data matters less, reputation matters more. It’s no longer enough to “be found”: you need to be chosen by the systems that answer on people’s behalf.
And there are no neutral results.
There never have been!
Algorithms always reflect the biases and intentions of their creators, while AI models amplify those same biases faster and more deeply. The algorithm builds the environment, but it’s human beings who feed it with signals: every click, share or interaction helps shape the perception of the world – and of brands – that the machine learns.
Yes, because the algorithm dictates the rules of the game, but it’s human behaviour that writes the match.
Those who can read and guide this relationship, understanding the invisible language that binds data, trust and reputation, will be visible to artificial intelligences too, and will be able to turn that visibility into sales, authority and brand strength.
Everyone else will be left standing at the starting line, spectators of a change that grants no extra time.
Thanks to Duane Forrester for this fruitful conversation, see you in the next episode!
