The Discovery Layer: what twenty years of getting businesses found taught me about AI search
Here is the rule I have traded on for more than twenty years, stated the way I would say it across a table. Search tools have one goal: get the right person to the right place. They are only good, and only get used, if they are accurate. They are not trying to mislead anyone. The methodology of how you feed the machine has changed over and over. The goal has never changed once.
Everything I know about visibility hangs off that rule, and I want to walk you through where it came from, because the road matters as much as the destination.
I started building websites and doing marketing in 2002 and founded my agency in 2008. In that first era, the basics of SEO set the winners ahead for over a decade. Not tricks. Basics: structure, clarity, saying plainly what you did and where. The businesses that learned the new feeding methodology early compounded for years while their competitors debated whether the internet mattered. That was the first time I watched the rule pay: the tool wanted to route people correctly, and it rewarded whoever made routing easy.
I also tested the rule from the wrong side, exactly once. When we first saw search spam, we tried it. It worked beautifully for a few weeks. Then Google shipped an update and it died, all of it, at once. We never touched blackhat methods again, and not because of ethics alone: the risk is simply not worth the reward for most businesses. But I want you to see what that episode actually proved. A system whose survival depends on accuracy must eventually correct anything that games it. I did not adopt my rule on faith. I tried to break it, and the system behaved exactly as the rule predicts.
The eras kept turning and the rule kept holding. When self-service tools started showing up, it was apparent where that went, and I got out of that business model. Then ecommerce, which mostly started going away too. Each shift was the same event wearing different clothes: the methodology changed, the goal did not, and the advantage went to whoever adapted their feeding first. Along the way the durable laws showed themselves. What is important is what everyone else says is important; self-description has never counted for much. And while links seemed like the game for years, what actually mattered underneath was being worth citing; the links were only ever a proxy for that. Today the agency business is really for companies that are already established, and nearly every field is hard for a new entrant to break into.
Which brings us to now, and to why I think this era deserves a name.
The platforms are deliberately moving discovery closer to the user's phone and further from your site. Where search once handed a person a list to evaluate, an AI now assembles the answer itself, before the person evaluates anything. There has always been a layer between the world's information and the person interpreting it, something deciding what is available to be seen at all. Editors did it, then shelves, then rankings, then feeds. Scholars have studied this gatekeeping for decades, so the observation is not mine. What is new is that the layer stopped showing its work. It no longer presents options. It presents conclusions. I call this the Discovery Layer, and understanding how it decides is, I would argue, the highest-leverage question in marketing right now, because what is available to be interpreted comes before everything a customer thinks or does.
I did not arrive at that framing casually. In 2024 I read every book I could get my hands on about AI. In 2025 I entered a doctoral program in media psychology on this exact territory, how information reaches people and becomes belief. And in 2026, the technology and the theory converged. The frameworks I now use come from working out, at a level deeper than tactics, how AI works, how information works, how it is transmitted, how trust is understood, and how psychology plays into all of it. I am telling you this because it explains the method of everything I publish: theory first, evidence graded in the open, tactics last, derived from the other two.
And here is the part that reframed the whole twenty years for me. This era is not just another turn of the wheel. In February 2000, in a keynote at Intel's Developer Forum, Larry Page described the ideal search engine as one that would understand everything on the web, understand what you wanted, and give you the right thing. Pressed on what that meant, he said it plainly: artificial intelligence, a search engine that could answer any question. That was the stated ambition a quarter century ago. Every era I lived through professionally was an approximation of it. The answer machine is not a detour from search. It is what search said it wanted to be before most of us were paying attention. And the psychology is why it will stick: an answer is simply easier than a search, and ease, as a rule, does not lose.
So that is the claim, and the road that produced it. The goal never changed. The methodology just completed its arc, from ten blue links to one assembled answer, and the layer that does the assembling now decides who gets found. The rest of what I am writing works out what that layer is made of, what the published evidence actually supports about how it selects, and what a business should do about it that pays off whether the current window stays open or closes. I have been wrong before, I have corrected myself in public before, and the whole project runs on the same rule as the machines it studies: it only works if it is accurate.
This essay is part of The Discovery Layer: How Businesses Get Found in the Age of AI, a book I'm writing in public. Each essay becomes a chapter, and the book expands them with the full evidence. Get on the list and I'll tell you when it ships.