76% of Google AI Overview citations once came from pages already ranking in the top 10 for the same query (Ahrefs, 2025). Eighteen months later Ahrefs re-ran it on a larger set: 38% (Ahrefs update). That single decline is the entire GEO-versus-SEO argument in one number.
GEO and SEO share most of their plumbing. Crawlability, indexation, page structure, credible sources, and topical depth all still count, because AI engines retrieve from the same web indexes that power classic search. What changes: AI answers cite passages instead of ranking pages, pull from beyond the top 10, and reward being mentioned rather than being clicked. The overlap is large and shrinking.
Two camps have staked out the extremes. One says roughly 80% of SEO carries straight over and GEO is a layer on top. The other, from practitioners like iPullRank’s Mike King, argues SEO “shouldn’t exist” as a separate discipline anymore. Both are describing the same data from opposite ends. Neither thinks fundamentals stopped mattering.
What carries over unchanged?
The gate work. If a bot cannot crawl your site, no amount of AI-era cleverness saves you. Robots rules, sitemaps, clean navigation, fast servers, and being present in the index all transfer directly, because retrieval is the first step of every AI answer and retrieval runs on an index you already know how to get into.
Topical authority carries too. The pages that earn AI citations are disproportionately the ones that earned rankings, even at the reduced 38% overlap. Credible, well-sourced content was the SEO advice a decade ago and it is the GEO advice now. The mechanism changed; the reward for being genuinely useful did not.
Schema and structured data carry over, and arguably matter more. Telling a machine exactly what your page is about was a soft ranking nicety. For an engine assembling an answer from fragments, it is closer to table stakes.
What stops working?
Rank tracking as your headline metric. In classic SEO you watched position one because position one took most of the clicks. AI answers do not hand out clicks the same way. The user reads the synthesized answer and often stops there. Watching your rank while your citations move independently is watching the wrong dial.
Single-keyword targeting stops working, because the engine fans your question out into sub-queries you never see. Optimizing one page for one head term aims at a target the model may skip entirely. If you want the mechanics of that, we covered how ChatGPT chooses which sources to cite.
And thin, high-volume content stops working faster than it did. An engine synthesizing an answer has no reason to cite your 300-word rewrite of a competitor when it can cite the competitor. The commodity middle got compressed. That was already true in SEO. AI answers just removed the participation trophy.
Where do rankings and citations part ways?
This is the part the overlap number hides. Ranking and citation are correlated, not identical, and the gap is where GEO becomes its own thing.
We watch this directly in our own scans. Running one site through two independently generated question sets, one source, fertilityclinicsabroad.com, showed up in 5 of 5 answers in the first set and 4 of 5 in the second. Its citation presence barely moved even as the surrounding brand mentions shuffled between runs. A source can own the citation layer of a category while individual rankings wobble underneath it. That stability is not something a rank tracker would ever show you, because a rank tracker is not looking at what the model actually said.
That is the practical divide. SEO asks “where does my page sit in a list.” GEO asks “how often, and how prominently, does the model repeat my brand when it writes the answer.” Those questions have different answers, and the second one is the one your board is asking.
So is SEO dead?
No, and anyone selling you that is selling you a rebrand. The honest read: SEO fundamentals are necessary and no longer sufficient. You still have to be crawlable, indexed, structured, and credible. You now also have to survive fan-out and get cited at the passage level, and you have to measure that with citation share rather than rank.
Treat GEO as the continuation of SEO with a new measurement layer, not a replacement for it. The teams that will struggle are the ones that either ignore AI answers entirely or throw out crawlability basics chasing a new acronym. Both are expensive mistakes in opposite directions.
The cheapest way to see where you sit is to stop guessing at rankings and read what the engines actually say about you. Our free scan runs a handful of real buyer questions through a live engine and shows you the citations it returns: see your answers.