SEO in the AI era: what still compounds and what just died

SEO built my career. Years of publishing grew a client's site a hundredfold, won first-place national rankings, and taught me how compounding works.

And a meaningful share of what I learned in those years is now dead. Not weakened. Dead. The honest move is to say which share, because most SEO advice being sold today was written for a search engine that no longer behaves the way it did.

The clicks that just died

Google now answers many searches itself, with an AI summary above the results, and the effect on clicks is measured.

Pew Research tracked real browsing behavior and found that when an AI summary appears, users click a result in 8% of visits, against 15% without one.

Roughly half the clicks, gone. And the links cited inside the summary itself? Clicked in 1% of visits. The reader gets the answer, the answer sits on Google, and the visit never happens.

Sit with what that kills. Every page whose job was answering a simple informational question is now competing with a free summary that loads first. Thin explainers, definition posts, volume content built to catch easy queries: that entire layer of SEO just lost its economics. If your traffic strategy was renting attention from basic questions, the landlord took the house back.

What still compounds

Now the half that survived, because the survivors share a trait: they are the things AI answers are built from, rather than the things AI answers replace.

Firsthand experience survived, and got more valuable. Google's own quality guidelines reward content that demonstrates lived experience, and the AI systems assembling answers need credible, specific, experience-backed sources to quote. Generic content feeds nobody; the case that went sideways and the pattern seen a hundred times cannot be summarized out of existence, because they exist nowhere else.

Branded search survived untouched. When someone types your name, no AI summary steals that click, because the searcher wants you, and the visit is the point. For an expert firm, the search that follows a referral or a talk was always the highest-value search you could win, and it still converts the same way. (What happens inside that search is the lookup.)

And earned citations survived. On that engagement I pitched a handful of sites, and the vast majority of the thousands of backlinks arrived from sites we never contacted. The same mechanism now runs through the machines: AI systems cite established, linked, consistent sources. Being the source the answer quotes is becoming the new first position.

Where I would put the effort now

For an expert firm, the reordering is clean. Publish the experience only you have, deeply, under one findable name. Win every branded search cold. Earn citations by being worth citing. And let the thin informational layer go without mourning it, because it was never sending you buyers anyway.

Notice what that list rewards: it is the same list that built authority before AI existed. The shift punished shortcuts, and it left the compounding assets standing. (That pattern holds across every tactic I have scored, in what works in marketing.)

What this scorecard will not tell you

Now the honest part, because you deserve the whole picture.

SEO timelines did not improve. The compounding assets still take months to move, and anyone promising fast rankings in the AI era is selling the same thing they were selling before it.

Traffic still is not revenue. My hundredfold client win came with the lesson that converting visitors into appointments stayed hard the whole way, and AI changed nothing about that half of the problem.

And this landscape is moving as I write. The numbers above are from 2025 research; the direction is clear, the specifics will keep shifting, and my scores shift with them. Check the practitioners before you bet, including against me.

The question that survived

Strip the acronym away and SEO was always one question: when someone goes looking, are you what they find?

AI changed who does the looking and where the answer assembles. It did not change the question. The experts who win the next decade of search are the ones whose captured experience is the material the answers get built from.

So look at what you have published. Is it the kind of material an answer gets built from? Or the kind an answer replaces?