Anton's Thesis

It's 2028. I just took the last two eggs from my AI smart fridge, and I'm all out.

Who will it buy new eggs from for me?

Doordash? Walmart? Instacart? Costco?

Hopefully my AI smart fridge knows that I am on a budget right now and need a good deal. I also need two cartons of eggs this week instead of one, since my friend is spending the week at my condo. I wonder if my AI smart fridge saw that in my calendar.

This will only be a hypothetical scenario for a very short period of time. I share it to present my hypothesis:

“For all of history, humans competed and adapted to discover more information. AI now forces information to compete and adapt to discover more humans.”

The first iteration of this is exemplified by the disruption that AI Answer Engines (AEO) have had on traditional web search and web economics.

For the last two decades, Google was the dominant interface for consumer discovery on the internet, controlling nearly 90% of global web search while operating one of the most profitable business models ever created, with Google Search itself estimated to generate margins approaching 75%. Entire industries emerged around this ecosystem: publishers created content, Google distributed traffic, advertisers monetized attention, and trillions of dollars in commerce flowed through the web as a result. But, facing the disruption of search by AI models, we are now watching Google disrupt the very model that made it one of the greatest businesses in modern history.

Will it be successful? Difficult to say.

For the first time in history, consumer search is fragmenting across more than one engine. People now take their questions to ChatGPT, Gemini, Perplexity, Claude, Grok, and Copilot, instead of clicking on 2-3 blue website links to find the product for them, they now ask follow-up questions to an AI model, which crawls dozens of websites each time. And every one of those frontier model providers is competing for the same thing: consumer attention and consumer queries, because that gives them the data, the customers, and the scale they need to win the age of AI.

To win that competition, each model has to deliver better recommendations, better answers, and better actions than every other model in the market. That creates a dynamic that most brands haven't woken up to yet: frontier models are actively prioritizing and weighing sources of information differently from one another as a deliberate strategy to build a competitive advantage.

What this creates is an environment where brands can no longer rely on a single playbook for consumer discovery. The algorithms and retrieval methods that matter are no longer the universal SEO principles that worked when Google owned the whole game. Which AI models your potential customers use, what questions they ask inside them, and the memory those models accumulate about them over time, all of it shapes what gets recommended and what gets ignored. The brands that understand this early will have an advantage that compounds over time. The ones that don't will lose ground they may never recover.

For any brand that never benefited from, or saw results from, Google SEO, now is your time to take advantage of this shift while the giants are sleeping.

The current wave of companies racing into this market will have to be incredibly nimble. Showing a company how they appear on ChatGPT for synthetically brainstormed questions that no one has actually searched and calling it a service will not hold up. That gets replaced the moment OpenAI, Perplexity, and others release their own versions of a Search Console, giving companies first-party data on how they actually appear for the questions their customers are genuinely asking. When that infrastructure exists, the gap between brands who acted early and brands who waited will become impossible to ignore.

The fundamentals of traditional search and what it takes to be recommended in AI answers are already diverging. Companies that wait to act risk losing something far more valuable than rankings: early representation across retrieval systems, citations, trust signals, and feedback loops shaping how frontier models learn to recommend brands. That advantage compounds early and becomes increasingly difficult to displace over time.

Today, AI models are still making recommendations using web infrastructure originally built for humans. But AEO is only the first transition layer toward a much larger shift, one where AI agents quickly become the dominant interface influencing consumer and enterprise purchasing decisions on behalf of people.

Voice search, visual search, memory-driven assistants, and agentic commerce will all rely on signals fundamentally different from the ranking factors businesses optimized for historically. As consumers place increasing trust in AI-generated recommendations over traditional advertising or influencers, their interactions with these systems become more personal, contextual, and conversational, reinforcing AI's role as the primary layer for discovery, trust, and decision-making.

From an economic perspective, the companies that adapt their information to be discovered, interpreted, and recommended by AI agents will fundamentally outperform the companies deprioritizing this transition.

With this, I'd like to conclude by returning to my hypothesis:

“For all of history, humans competed and adapted to discover more information. AI now forces information to compete and adapt to discover more humans.”