The SEO Fallacy

To survive, you must abandon the crowded center of consensus and instead focus on unique ideas and questions no one else is asking.

One of the most common tools in the SEO toolbox was known as the “Skyscraper Technique.”

The premise was logical, almost scientific. If you wanted to rank for a competitive term like “Project Management Best Practices,” you would look at the current number one result on Google. If that article had ten tips, you would write an article with twenty tips. If their article was 1,500 words, yours would be 3,000. You would add better graphics, more current data, and a catchier headline. You would build a taller building, and Google, recognizing your superiority, would crown you the winner.

This strategy worked because Google was a sorting engine. Its job was to rank discrete documents. It compared Document A to Document B and decided which one was “better.”

But in the age of synthesis, the Skyscraper Technique is ineffective.

When an LLM answers a user’s prompt, it is not looking for the best document. It is looking for the consensus answer. It is scanning the skyscrapers, the two-story houses, and the shacks, and it is calculating the statistical average of what “Project Management” looks like.

If you write a 5,000-word definitive guide that perfectly mirrors the consensus view of your industry, you haven’t built a moat. You have simply contributed a very high-quality brick to the wall of noise that the AI uses to train itself.

The Mechanism of Mediocrity

To understand why “better” doesn’t beat AI, you have to respect the underlying math of the Large Language Model.

At their core, these models are prediction engines designed to generate the most probable next token (word) in a sequence. They are trained on the internet, which means they are trained on the average of human expression. When you ask an AI a question, it is statistically incentivized to give you the most likely answer, the answer that aligns with the majority of its training data.

This means LLMs are, by definition, engines of conformity. They smooth out the edges. They remove the outliers. They regress to the mean.

If you are a content creator following traditional SEO advice, you are likely using tools like SEMrush or Ahrefs to find “keywords with high search volume.” You see that 50,000 people a month search for “How to write a cover letter.” So, you write a guide on cover letters. You look at what’s already ranking and you make sure to cover the same points, because that’s what Google expects.

By doing this, you are explicitly training the AI to ignore you. You are telling the model, “I agree with everyone else.”

When the AI synthesizes an answer for the user, it will output the generic, consensus advice. It won’t cite you because you didn’t say anything unique. You were just part of the choir.

The Canary in the Kitchen: The Recipe Apocalypse

If you want to see the future of your industry, look at what happened to recipe blogs.

For years, recipe bloggers were the masters of the SEO game. They knew that to rank for “Chocolate Chip Cookies,” they couldn’t just post a recipe. Google required “dwell time” and “content depth.” So, bloggers wrapped their ingredients in 2,000-word narratives about their grandmother’s farmhouse, the smell of autumn, and the emotional resonance of melted chocolate.

Users hated it. They mocked the “life story before the recipe.” But it worked. It drove traffic, which drove ad impressions, which drove revenue.

Then came the AI.

If you ask ChatGPT for a “chewy chocolate chip cookie recipe,” it does not give you a story about autumn. It gives you the ingredients and the instructions. It extracts the data (the recipe) and discards the wrapper (the narrative).

The value of the blog post was the data. The wrapper was just friction. The AI removed the friction, and in doing so, it destroyed the business model. The user got the cookie, and the blogger got zero clicks.

This is the Great Flattening. The AI separates the insight from the creator. If your insight is generic (flour, sugar, butter), you are anonymous.

The Coder’s Dilemma

The developer community saw this same dynamic unfold in real time, and the numbers tell a brutal story.

For over a decade, Stack Overflow was the oracle of programming knowledge. If you had a coding problem—a cryptic error message, a tricky algorithm, a framework you couldn’t quite grasp—you had a ritual: search Google, land on a Stack Overflow thread, scan for the green checkmark, copy the solution.

This ecosystem thrived on a simple exchange. Developers contributed their expertise for free, building reputation through upvotes and accepted answers. In return, Stack Overflow captured massive traffic, which it monetized through ads, job listings, and enterprise products. The contributors got status. The platform got revenue. The searchers got solutions.

Then came GitHub Copilot and ChatGPT.

Now, when a developer needs to write a function, they don’t search. They don’t even leave their IDE. They simply describe what they want—“write a Python function to reverse a string”—and the AI generates the code instantly. The AI, having been trained on millions of Stack Overflow threads, synthesizes the consensus solution without attribution, without a visit, without a click.

Traffic to Stack Overflow didn’t just decline. It collapsed. Between 2022 and 2024, the site saw a 50% drop in new questions posted and a corresponding nosedive in page views. The contributors who had spent years building their reputation scores—the “10k rep” users who were the platform’s aristocracy—suddenly found themselves ghostwriting for machines that would never credit them.

Why? Because the “answer” was a commodity.

The code to reverse a string in Python is always the same. It doesn’t matter who wrote it first, who explained it best, or who has 50,000 reputation points. The AI doesn’t care about expertise signals. It cares about statistical patterns. It reads every solution, averages them out, and outputs the consensus.

Stack Overflow was a cathedral built on consensus knowledge. And consensus is undoubtedly what AI consumes without attribution.

The contributors weren’t providing unique insight. They were documenting standard practice. They were building the corpus that would make their own platform obsolete. Every well-formatted answer, every carefully explained algorithm, every “here’s how you do this in Python 3.8” was training data for the model that would eventually replace the need to visit Stack Overflow at all.

This is the paradox of commodity expertise: The better you document the consensus, the more efficiently the AI can extract it.

And here’s the kicker—Stack Overflow’s model was designed to create consensus. The voting system promoted the most popular answers to the top. Duplicate questions were closed and redirected. Unconventional solutions were downvoted as “not idiomatic.” The entire platform was a consensus-generating machine.

Which made it the perfect training ground for an AI that specializes in outputting the most statistically likely answer.

The Trap of High Volume

This leads to a terrifying conclusion for modern marketing: Targeting high-volume keywords is now a liability.

In the SEO era, high volume was the prize. It meant a large addressable market. It signaled commercial intent and justified budget allocation. You could point to a keyword getting 50,000 monthly searches and make a compelling business case for creating content around it.

But in the AI era, high volume implies high consensus. If everyone is searching for it and everyone is writing about it, then the AI has already mastered it. The corpus is complete. The statistical average is locked in.

If you try to compete for “CRM software,” you are fighting a losing battle against a machine that has read every review of Salesforce and HubSpot ever written. The AI can synthesize a comparison table in seconds that is more objective, more comprehensive, and more up-to-date than your blog post could ever be.

You cannot win by answering the questions that everyone is asking. You cannot win by providing the answers that everyone expects.

This is the Great Flattening in action. The more common the question, the more commoditized the answer. The more traffic a keyword drives today, the less value it will have tomorrow.

The SEO playbook told you to climb the mountain toward consensus—find what people search for, see what ranks, and make something “10x better.” But consensus is precisely what makes you extractable. You were optimizing for a game that no longer exists.

From Volume to Scarcity

So what do you optimize for instead?

Idea scarcity.

To survive the Great Flattening, you must abandon the crowded center of the bell curve. You must run toward the edges—not because the edges are more comfortable, but because they’re the only place where attribution still matters.

You must stop trying to build a taller skyscraper in the city center and instead build a fortress on an island that isn’t on the map yet. You must stop chasing the questions everyone is asking and start defining the questions no one knows to ask.

This requires a fundamentally different discipline:

Observation over optimization. Instead of analyzing keyword difficulty scores, you must go into the field. Talk to customers. Watch how they work. Notice the friction points they’ve learned to accept as “just how things are.” The Tuesday Lag wasn’t found in an SEO tool—it was found in a warehouse.