How typically are pages that rank within the SERP cited in AI Overview?
Final 12 months, we analyzed 1.9 million citations to reply that query.
However like all the things AI-related, so much has modified since then.
For instance, as of January 2026, AI will appear like this: Powered by Gemini 3 Beforehand, we relied on earlier Gemini fashions, together with the two.5 variant, which higher answered long-tail searcher questions.
Since our final research, we have additionally improved the best way we analyze Ahrefs, permitting you to see extra of the citations that seem in your AI overview.
So we determined it was time to replace our findings to raised reply the unique query.
This time, we analyzed 863,000 key phrase SERPs and a complete of 4 million AI abstract URLs, greater than double the earlier evaluation.
All this knowledge comes from Ahrefs Model Radar, an AI visualization software.
Right here’s what we discovered…
In our evaluation, we checked out instances the place the identical URL appeared in each the AI overview and common SERPs. similar question.
The primary take a look at checked out all the outcomes returned on the web page. Contains the standard blue hyperlinks and SERP options.
For instance, we tracked advertisements, featured snippets, PAA bins, and video packs. and Natural listings as separate blocks.
This exhibits that 37.9% of the URLs cited within the AI overview seem throughout the first 10 blocks.
The remainder was cut up nearly evenly between ranks 11-100 (31.2%) and high 100 blocks and past (31.0%).
The second take a look at centered solely on normal blue hyperlinks and ignored advertisements and different SERP options. Ultimately, the distributions have been nearly related.
| natural place | Share of cited URLs |
|---|---|
| Ranked in high 10 | 37.10% |
| Rank 11-100 | 26.20% |
| Not ranked within the high 100 | 36.70% |
The one slight distinction was that lots of the pages cited within the AI overview fell exterior the highest 100.
This means that AIO might cite pages that don’t seem in common search outcomes, however seem in SERP options.
So what does this knowledge really imply?
Effectively, no less than in comparison with our unique research, we are able to see that Google selects far fewer pages instantly from the unique SERP (about 76% in July 2025 vs. about 38% now).
To me, this means that AI Overview depends extra on sources showing in fan-out question SERPs than on direct search outcomes.
Google confirmed Each time a person searches and the AI is triggered, the system performs a “question fanout.”
That is when the preliminary question is cut up into a number of associated subqueries.
Probably the most regularly considered pages inside these subquery SERPs are cited within the AI Overview.
Subsequently, with the arrival of Gemini 3, fan-out queries might play a bigger position in supply choice.
You could increase your question extra aggressively than earlier than, or chances are you’ll increase your question extra broadly by pulling from associated SERPs which have fewer pages rating on your unique question.
Our analysis revealed some very fascinating issues about YouTube.
Pages cited in AI Overview that didn’t rank Prime 100 outcomes on Google For a similar key phrase, 18.2% have been YouTube URLs.
And these YouTube URLs accounted for five.6% of all cited AI overview URLs throughout the dataset.
In different phrases, YouTube accounts for a good portion of complete AI overview citations, regardless that it doesn’t rank within the SERPs for direct queries.
Based on knowledge from Ahrefs Model Radar, YouTube is at present essentially the most cited area in AI Overview, rising 34% previously six months.
Our different analysis helps the significance of YouTube in AI visibility.
For instance, our analysis throughout 75,000 manufacturers revealed that mentions in video titles, transcripts, and descriptions on YouTube are the elements most strongly correlated with AI overview visibility.
With all this in thoughts, it is smart to make YouTube part of your AI search technique.
See which YouTube movies point out your model with Model Radar.
You too can see which movies have been pulled into AI Overview by checking the “Cited in AI” column.
If rating in the identical SERP as AI Overview isn’t sufficient to get an AIO quotation, what’s?
I just lately wrote an article like this How to rank in AI Overview.
In it, I talked about how search engine optimization is shifting from optimizing content material for key phrases to optimizing for fan-out queries.
search engine optimization advisor, Ethan Razukwrites about precisely this in his glorious article Google’s Query Fanout Technique and What SEOs Need to Know About It:
If you can also make textual content in your paperwork (or different media) extra related to your fan-out queries, you might have a greater probability of getting mentions and citations in your AI-generated solutions. So with regards to how question fan-out impacts your rankings, know that we’re optimizing your complete person journey, not particular person key phrases, and fan-out queries information that course.
The issue is that Google would not publish the simulated fan-out queries it generates for every AI response.
And even when they did, the define of AI is probabilistic.
Their content material, citations, entities, and fanouts change with every question. To be taught extra, learn the research, “The profile of AI modifications each two days (however it by no means modifications its thoughts).”
Nonetheless, there are methods you’ll be able to replicate the question fanout course of your self utilizing instruments resembling:
- Kuforia from iPullRank founder mike king
- this Gemini API + Screaming Frog Workflow from go fish digital co-founder dan hinkley
- AI visualization fanout by word drift
No matter route you select, it’s good to be sure to cowl the subject from all angles.
There are just a few extra methods to do that utilizing Ahrefs.
Cowl all of the fundamentals utilizing the mum or dad subject
Use it to search and cluster content around related themes, not just single keywords.
Investigate fan-out queries generated by other AI tools
Fanout queries are simulated and change from run to run.
It is not so important whether you can perfectly reflect Gemini’s internal expansion.
Even more important, many extended examples can be analyzed at scale to identify recurring themes, topics, and angles of intent that the AI may consider when building answers.
In the AI Responses report, you can explore millions of examples of ChatGPT and Perplexity fanout queries associated with both database prompts and your own custom prompts. brand radar.
Create inclusive content with AI Content Helper
Our team designed AI content helper that allows you to optimize your content specifically for AI search.
Generate relevant fan-out queries and measure the cosine similarity between the topics covered in your content and the topics your SERP or AI responses may be trying to address.
As you write, colored highlights appear to show you how thoroughly you covered each topic.
You can use this to benchmark your content against the top 10 results in the SERPs, but as this very study revealed, in many cases actually Competing companies don’t necessarily rank on page one. Sometimes it ranks even lower, sometimes it shows up in a rich SERP feature, and sometimes it gets picked from a completely different SERP.
So now you can use Ahrefs’ AI Content Helper to benchmark your content against your own custom competitors, including those that regularly earn AI Overview citations.
See AI Overview Quotes on Brand Radar
Studying your own AI Overview citations can help you claim more visibility because you’ll see exactly what the system already trusts you with, and which parts of the neighborhood are not fully picked.
For example, you can:
- Seeing which pages and topics have already been cited will give you an idea of what Google considers you to be authoritative.
- Instead, identify gaps where competitors are cited and reverse engineer what additional contexts, entities, or subtopics they cover.
- Identify formatting patterns (definitions, statistics, step-by-step, video embeds) that tend to draw you in to AI answers.
- Boost pages that are ranked but not cited. It often indicates that the content or clarity of the topic is not fully aligned with the extended intent.
In other words, citation analysis moves you from being “ranked” to being “selected as a source,” but this is not necessarily the same thing.
final thoughts
The landscape of AI is changing, and rankings for a user’s exact queries no longer guarantee visibility.
If you want to understand what Google actually trusts and how to get more citations, start by studying patterns in your data.
As always, if you have any questions about this study, please feel free to email me. linkedin.
