"If you rank well in Google, you increase your likelihood of being cited by AI. Strong SEO is still the foundation of AEO." That's Lily Ray, on stage at Affiliate Summit West 2026, making a claim that would have been uncontroversial a year ago. It's more complicated now, and the data that complicates it comes from the same company most SEOs already trust.
The claim, on stage
It's worth being precise about what actually got said, since keynote lines have a way of getting flattened into something punchier once they leave the room. Ray wasn't arguing that AI citation is identical to ranking, or that a GEO program is redundant if a site's SEO is already strong. The claim was narrower and more defensible than that: ranking well increases the likelihood of citation. A likelihood claim, not a guarantee, is a much easier position to hold up against contrary data than an absolute one would be, and it's worth remembering that distinction before treating the Ahrefs numbers as a rebuttal rather than a refinement.
Ray's position isn't a hot take. It's the mainstream, structurally-grounded view: large language models running retrieval-augmented generation pull from live search results, so a page's organic ranking performance feeds directly into whether it gets pulled into an AI-generated answer. That's how the systems are built. A page Google's own ranking algorithm has already decided is the best answer to a query is, all else equal, a reasonable candidate for an AI system drawing on the same underlying index.
It's also the version of the story that keeps SEO and GEO pointed in the same direction, which is exactly why it's an appealing thing to say from a keynote stage. If ranking still drives citation, then a decade of link building and technical SEO investment isn't stranded, it's the foundation the AI-visibility layer gets built on top of. Nobody wants to be the one telling a room full of SEOs that the skill they've spent years building stopped mattering.
Ray isn't a hype figure making that argument for stage effect, either, which is worth being clear about. She's spent the past two years publishing some of the more skeptical, data-first work in the field, including her own research showing self-promotional listicles cited in AI Overviews frequently lose the actual recommendation to a named competitor. Someone with that track record telling a room that SEO fundamentals still drive AEO carries noticeably more weight than the same exact line from someone simply selling a GEO retainer off the back of it.
The data, eight months later
Ahrefs first measured the relationship between organic rank and AI Overview citation in July 2025: 76.10% of cited pages ranked in the traditional top 10, with the median primary cited URL sitting at position 2. That's about as clean a confirmation of Ray's claim as you could ask for. Rank well, get cited, in three out of four cases.
| METRIC | JULY 2025 | MARCH 2026 |
|---|---|---|
| Cited pages also ranking top 10 | 76.1% | 38% |
| Cited pages ranking positions 11-100 | 9.5% | 31.2% |
| Cited pages not ranking in top 100 | 14.4% | 31.0% |
Eight months later, Ahrefs re-ran essentially the same analysis at much larger scale, 863,000 keywords and 4 million AI Overview URLs, and the top-10 overlap had roughly halved, down to 38%. The share of citations coming from pages ranking outside the top 10 more than tripled. That's not a small drift. It's the single biggest reported swing in this specific relationship since anyone started measuring it, and it happened inside less than a year.
Ahrefs itself flagged one caution worth repeating rather than glossing over: its citation-detection methodology improved between the two studies, which means part of the gap could reflect better measurement rather than a purely behavioral shift in how Google's AI Overviews select sources. That caveat matters, and it doesn't erase the finding. Even generously assuming some of the movement is a measurement artifact, a swing this large, corroborated by Google's own public explanation of how fan-out queries work, is not plausibly explained away entirely by better detection tooling.
Why both things are true
The mechanism Ahrefs points to is fan-out. When an AI Overview triggers, Google doesn't just answer your literal query, it silently expands it into a cluster of related sub-queries and pulls citations from whichever pages perform best across that wider cluster, not just the original search term. A page can rank #1 for the exact keyword you're tracking and still lose a citation slot to a page that ranks lower, or doesn't rank at all, for that keyword but happens to rank well for one of the sub-queries Google generated behind the scenes.
This is why the two findings can both be true without either side backing down. Ray's framing is about SEO as the necessary foundation: without organic ranking signal somewhere in the picture, an AI answer engine built on retrieval-augmented generation has nothing to retrieve. Ahrefs' framing is about SEO as a sufficient predictor for one specific page against one specific query, and on that narrower question, the data has moved a lot in eight months. Foundation and sufficiency are different claims. Conflating them is how this debate turns into two camps talking past each other instead of one accurate, if slightly more complicated, picture.
“A top ranking no longer guarantees a citation. It's still the strongest single predictor available. Those are two different sentences, and the data supports both of them at once.”
That's the reconciliation. Ray's claim is about ranking as a predictor in aggregate, and on that measure she's correct: the primary cited URL in Ahrefs' own data still tends to rank well, and no other single signal comes close to organic rank as a predictor of citation likelihood. The 38% figure isn't a rebuttal of that. It's a measurement of something narrower and more specific: how often the exact page ranking top-10 for your tracked keyword is also the page that gets cited, as opposed to some other page from the same domain, or a competitor's page, that ranks for a sub-query you were never tracking in the first place.
Put those together and the honest version of the claim looks like this: ranking still matters enormously, but the unit of what needs to rank has changed. It used to be one page against one keyword. Increasingly, it's a cluster of content against a cluster of sub-queries you can't see in any keyword tool, because fan-out queries mostly don't have search volume attached to them at all in the tools built to plan around them.
There's a version of this shift that should worry a link building team more than the aggregate numbers let on, too. The 31% of citations now coming from pages ranking outside the top 100 entirely means Google, in a meaningful minority of cases, is citing pages that a traditional SEO audit would have flagged as essentially invisible. Those pages didn't get there through classic ranking signals at all; they got there by answering a sub-query cluster well enough that fan-out retrieval found them regardless of where they sat in the primary SERP. That's not noise in the data. It's a second, parallel path to citation that link building strategy hasn't historically had a name for.
What this changes for link building
For link building and content strategy specifically, this argues against the instinct to concentrate all authority-building effort on a single hero page targeting a single head term. That page can win its own keyword cleanly and still lose the citation to a competitor's page answering a sub-query fan-out generated, one that never shows up in a rank tracker built around your target keyword because it was never the target keyword in anyone's plan. The topical authority research pointing toward depth-across-a-cluster over isolated page-level optimization is the same finding, arrived at from a different angle.
It also raises the value of earning links and mentions across a full topic cluster rather than funneling every backlink at one canonical page. A domain with corroborating authority spread across the adjacent questions a buyer might ask, not just the one you optimized for, is positioned to catch citations from sub-queries you never explicitly planned for, precisely because you can't see most of them in advance. That's a genuinely different brief for a link building program than "build authority to page X": build authority to the cluster, and trust that individual pages within it will pick up citations for sub-queries that don't have names yet.
This is also, worth saying plainly, exactly the kind of shift an agency built on link building alone would rather not talk about, and exactly the kind a GEO-literate one should be leading the conversation on. It changes reporting, too, not just strategy. A link building program judged purely on whether its target page holds the #1 spot for its target keyword is measuring exactly the metric that's become least reliable as a citation predictor. A program judged on whether the domain's cluster of related pages collectively rank well across a topic, with links and mentions spread proportionally rather than concentrated on one hero page, is measuring something closer to what actually determines fan-out citation odds today. Clients asking why a page can hold position one and still lose visibility in AI answers deserve that distinction explained in exactly these terms, not a vague gesture at "the algorithm changed."
None of this makes Lily Ray wrong. It makes her claim true at the altitude she made it, and incomplete at the altitude link builders actually operate day to day. Rank well and you're still the favorite. Rank well on only the one keyword you were tracking, and you're increasingly playing a game whose real scoreboard, the fan-out cluster, was never visible to you in the first place. Ahrefs published the full methodology and dataset behind the updated figures directly, and it's worth reading in full before deciding how much weight to put on any single number pulled from it: the Ahrefs update on AI Overview citations covers both the July 2025 and March 2026 measurements side by side.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.