AEO is not SEO with a different acronym — and most tooling gets this wrong

01Crawl, rank, click. Crawl, retrieve, synthesize.

The SEO pipeline exists to produce order. Googlebot crawls a page, the index files it as a discrete URL, and a ranking pass weighs upward of 200 signals, still descended from the link graph, into a single ordered list for a given keyword. Win the order and you win the click. Every rank tracker, every SERP-position dashboard, every AEO tool that just bolted a "ChatGPT visibility" tab onto an existing SEO product exists to monitor that one list.

The AEO pipeline doesn't produce a list at all. A page gets crawled, then chunked into passages and embedded into a vector index, so the unit of competition drops from the page to the paragraph. When a query lands, most answer engines don't run it once. Google's AI Overviews split a question into several related sub-queries, a technique known as query fan-out, then retrieve passages across the SERPs those sub-queries generate. A model reads whatever comes back and drafts an answer, choosing which two or three passages earned a citation. The same page can win one sub-query's retrieval pass and lose the next, a behavior a single keyword rank was never built to describe.

02What a retriever actually pulls

Retrieval scores a passage on semantic closeness to the sub-query in front of it, plus a handful of freshness and authority signals. Backlinks aren't in that scoring function. That's the mechanical reason a page sitting 60th on Google can get cited by an answer engine while the page holding position 2 gets skipped entirely.

Ahrefs's March 2026 update to its AI Overview citation study makes the gap concrete. Across 863,000 keyword SERPs and 4 million citation URLs, only 37.9% of the pages cited in AI Overviews also ranked in Google's own top 10 organic results. 31.2% sat in positions 11 through 100. The remaining 31.0% weren't in the top 100 blocks at all, pages a rank tracker would call invisible. Seven months earlier, the same team had put the top-10 overlap closer to 76%. The two pipelines used to look correlated, mostly by coincidence. Ahrefs's own read on why they're drifting apart now: once a question gets split by fan-out, the page that wins depends on which sub-query it happens to answer well, not on its rank for the original phrase.

03What synthesis keeps, and why it isn't your best page

Retrieval decides what's in the room. Synthesis decides what gets quoted, and that step runs on a different set of preferences again. The closest thing to a controlled study of it is a 2024 paper out of Princeton, Georgia Tech, IIT Delhi, and the Allen Institute for AI, titled "GEO: Generative Engine Optimization" and presented at KDD. The team built a 10,000-query benchmark across eight domains, tested nine content interventions against a Bing-Chat-style engine, then validated the strongest ones directly against Perplexity.

The interventions that moved the needle weren't ranking factors. A clear statistic dropped into a passage helped. So did a quotable line, and so did a shift to a more authoritative register. Together, they lifted citation rates measurably enough that the paper's own abstract reports gains of up to 40% in generative-engine visibility. A page with a clean, quotable number in paragraph four can out-cite a page that outranks it by fifty spots. The model is grading the sentence it's about to lift.

04Why the rank tracker is lying to you

Rank tracking still matters. Google's blue links send the majority of the traffic most sites live on. But it's the wrong instrument for a pipeline that was never ranking anything to begin with. A content team watching only the classic dashboard can see a page sit at position 3 for months, flagged stable, while its citation share on a second engine quietly falls to zero because a competitor rewrote a paragraph with a citable number in it. The dashboard has nothing to say about that; it was never watching for it. Retrieval cares only whether a passage showed up for the sub-query in front of it, this one time, for this one answer.

05What we track instead

Caposeo was designed around that split from the start. It watches the same query land on Google, ChatGPT, Perplexity, Claude, and Gemini at once, 12.4K keywords deep right now, and rolls crawl coverage, Core Web Vitals, structured data, and citation share, not link equity, into a single health score, currently 87 for the sites we watch most closely. Three of those four inputs describe whether your content is retrievable at all. The fourth describes whether it got quoted once it was.