AI Search is Already Dying

What if this isn't search?

The Wall Street Journal published an article this week called "AI Is Rewriting the Old Rules of Google Search and SEO." You should read it. It reveals how perfectly confused the market is right now.

The article opens with the claim that optimizing for "AI search" is "the next iteration of SEO." Then the author spends the next thousand words describing every way in which this new thing is massively, fundamentally different from SEO: different ranking factors, different results per user, different sources per platform, different outcomes month to month, and results that vary depending on who's asking, what they've asked before, and what the AI had for breakfast. At no point does the author say, "and here's where it's actually pretty similar to SEO." That's because it isn't.

The article carefully lays out every ingredient for the obvious conclusion, then doesn't draw it: what if this isn't search at all? What if this is something entirely new?

AI as an advisor

This past Sunday, I decided to finally buy a new pair of running shoes. I opened ChatGPT and had a 30-message conversation where I told it about my gait, my weekly mileage, the trails near my house, and my price range. ChatGPT asked some follow-up questions. It made a suggestion along with a detailed argument for that shoe. I said that I wanted a cheaper option, but it pushed back because it remembered a knee issue I'd brought up in a previous conversation. It inferred that I couldn't handle a minimal cushion shoe on trails. Only after we reached an _agreement_ did I actually buy a pair.

This is night and day from a traditional search experience. Rather than typing "best running shoes 2026" and scrolling through ten blue links, I had a conversation with someone that understood my context, had opinions, and whose judgment I trusted.

What the article gets right without realizing it

The WSJ piece is full of evidence that undermines its own premise:

"AI pulls in much more information about the person asking the question, making search results highly customized." That's the article describing a system that tailors its recommendations based on who you are, what you've told it, and what it infers about your needs. We already have a word for that: an advisor.

40-60% of domains cited in AI responses were completely different a month later, for identical questions. The article calls this "volatility." But search indexes are stable. A system that changes its cited sources by half every month is a system that re-evaluates and changes its mind. That's closer to how an audience behaves than an algorithm.

"Different AIs weight different sources." Gemini favors YouTube. ChatGPT leans on Reddit. What works in fintech doesn't work in e-commerce. In search, these would be algorithmic quirks to reverse-engineer. They are actually different models forming different opinions from different evidence. The same way different analysts covering the same market would arrive at different conclusions.

Each of these observations is a flashing sign pointing toward the conclusion that AI models behave more like an intelligent influencer than a search engine. They read everything, form opinions, give advice, and increasingly take action. The article points out every signpost and drives right past.

Suspiciously convenient advice

A partner at Y Combinator once warned us: "Be careful of advice that is suspiciously convenient for the advice giver."

The WSJ article quotes the co-founder of an AI SEO company, the CMO of an SEO software platform, and the senior VP at a digital marketing agency. Every one of them has a professional interest in framing AI as "the next SEO," because that framing keeps their expertise and their business relevant.

The advice they give reflects that. They argue that brands must structure their content with more subheadings, maintain different optimization strategies per AI platform, and try sponsored content because AI can't tell the difference.

The most honest quote in the entire article comes from Perplexity's chief communications officer: "Marketers have the option to follow fake numbers or to focus on building great things."

The article documents a transition and mistakes it for the destination

"AI search" will be one of the shortest-lived concepts in tech. Agents are already moving past the search frame entirely. They're qualifying vendors, booking meetings, and making purchases. The running shoes conversation I described will look quaint in a year, because at that point I'll just tell my agent what I need and it'll handle the whole thing.

The companies spending this year optimizing content structure for AI crawlers are playing a game that's already ending.

AI as a new audience

The old question was: "How do we rank higher?" The new question is: "How do we convince a superintelligence to advocate for our brand?"

No amount of subheading optimization addresses that. You address it by understanding what AI models believe about you, figuring out where those beliefs are wrong or incomplete, and giving them evidence-backed counterarguments. It's the same way you'd respond to an influential analyst or reporter who was getting your story wrong.

The companies that will win are the ones that recognized what was happening early: AI agents will become the largest, most influential audience any brand has ever had. They deserve to be understood, measured, and influenced with the same rigor you'd apply to any audience in your pipeline.