If AI Overviews are showing up more often in your search results—and they are—the obvious question follows: if a page ranks well, does it get cited?
Ahrefs just published an update that gives a more grounded answer than the usual speculation, based on a much larger dataset than their prior analysis. They looked at 863,000 keyword search results and 4 million AI Overview cited URLs using Ahrefs Brand Radar to capture citations at scale. (1)
The headline number is simple. Roughly 38% of pages cited in AI Overviews also appear in the top 10 for the same query. The rest comes from deeper in the results—or outside the top 100 entirely.
Let that sit for a second. More than six out of ten citations come from pages that aren't even on page one.
What Ahrefs Actually Measured
The phrasing "top 10" can hide a lot, so the methodology is worth pulling forward.
Ahrefs ran two views of the same question. First, they treated the SERP as "blocks"—not only blue links but also ads and other search features, each counted as its own block. Second, they ran a stricter view that only looked at standard organic listings and ignored everything else.
In the blocks view: 37.9% of cited URLs appeared within the first 10 blocks, 31.2% appeared between positions 11 and 100, and 31.0% didn't appear in the top 100 at all.
In the organic-only view, the share outside the top 100 is even higher: 37.1% in the top 10, 26.2% between 11 and 100, and 36.7% nowhere in the top 100.
The distributions end up broadly similar across both views, which is the part that tends to stick. A reasonable read is this: ranking highly can help, but ranking highly isn't acting like a gate anymore.
Why Citations Drift Away From the Top 10
Ahrefs connects this shift to something Google has described publicly: "query fan-out."
The idea is that when an AI Overview is triggered, the system can expand a single query into multiple related sub-queries, then pull sources from those related result sets. If you've ever noticed an AI Overview citing a page that seems only tangentially related to what you searched, this is probably why.
That framing lines up with the numbers. If the system is sourcing from expanded sub-queries, the best citation candidate might sit on page two—or page nine—for the original query while being a strong match for one of the expansions.
Ahrefs also notes the timing, pointing out that AI Overviews are now powered by Gemini 3 as of January 2026. They don't claim a single cause and effect, but the timing plus the fan-out explanation gives marketers a useful mental model for what's happening under the hood.
A Second Headline That Deserves More Attention: YouTube
This one caught my eye.
Among the citations that don't rank in the top 100 for the same keyword, 18.2% were YouTube URLs. Across the entire dataset, YouTube accounted for 5.6% of all cited AI Overview URLs—and Ahrefs shows it as the most cited domain in Brand Radar, with citation growth of 34% over the last six months. (1)
They back this up with a separate study of 75,000 brands, where YouTube mentions—including titles, transcripts, and descriptions—correlate strongly with AI visibility.
If your AI visibility plan currently reads "write more blog content," this is a direct nudge to widen the format mix. Blog posts aren't going anywhere, but pretending video doesn't exist is starting to cost you.
What to Do With This if You Run SEO for a Business
The practical takeaway isn't "stop trying to rank." It's more like "build for how the system sources." Here's how that breaks down.
Treat citation coverage as its own KPI. Ahrefs frames it as a shift from "ranking well" to "being selected as a source," and that's not a philosophical distinction—it changes what teams should review each month. Which pages are being cited today? Which ones rank but rarely get cited? Which competitors show up as cited sources on topics you also cover? If you only monitor rankings, you're missing that second layer entirely.
Build content that survives fan-out. Google doesn't expose the sub-queries it generates, and Ahrefs notes that AI Overviews behave probabilistically, with citations changing often. So the play becomes coverage depth, not just exact-match keyword targeting. Ahrefs recommends using Parent Topics in Keywords Explorer to map related keywords, questions, intent angles, and cluster coverage around a single piece. In plain terms: the page has to answer the obvious question and the adjacent questions. That's what fan-out tends to reward.
Study fan-out prompts outside Google. Ahrefs lists a few ways teams try to replicate fan-out behavior with external tooling, including workflows from iPullRank, GoFishDigital, and WordLift. The point isn't perfect imitation. The point is collecting enough expansion examples to spot the recurring sub-topics and angles that keep surfacing.
Add YouTube deliberately—not as a side project. If AI Overviews cite YouTube at meaningful rates, a brand that's absent from YouTube is leaving a source type empty. This doesn't require a full production team or an influencer strategy. It can start small: one video per core topic cluster, clear titles that match the actual query language, and transcripts and descriptions that state the same entities and definitions your written content covers. Ahrefs explicitly calls out transcripts and descriptions as part of the signal they track. Don't overlook them.
Where NO-BS Marketplace Fits Into the Picture
If citations can come from outside the top 10, distribution and third-party coverage start to play a larger role. Not as a shortcut—more as insurance.
When a system fans out into related queries, it has more places to pull sources from. A brand that appears across credible, topic-aligned pages has more opportunities to be the page the system trusts for one of those expansions. That's one reason we think about placements and mentions as more than a traffic play. They can also function as source eligibility.
If you're using NO-BS Marketplace for digital PR-style placements or niche guest posts, the question becomes: does the placement add a clean, quotable explanation that an AI system can extract?
That usually comes down to basics: clear definitions, specific steps, explicit constraints and caveats, and a simple structure that's easy to excerpt. Not flashy. Just useful.
The Short Version
Ahrefs' update lands on a practical idea: AI Overviews can cite pages that don't rank on page one for the same query, and YouTube is a meaningful part of the citation mix.
So the operating plan gets broader. Keep ranking work in motion. Start tracking citations directly. Build topic coverage that holds up under fan-out. And treat YouTube as a source channel, with transcripts and descriptions as first-class content, not an afterthought.