The majority of the advice we are getting for AI overviews is just speculation, so I analyzed multiple AI overview responses plus deeply reviewed third-party research and the studies for Google Guidance towards finding out what works and what doesn’t. Through different studies, this shows that AI overviews have caused a massive drop for CTR (click-through rate) at around 34.5%, with the new research done by Seer Interactive that reports a drop that can go as high as 61%.
And the massive research coming out from Pew Research shows the users that can encounter the AI summary clicks for traditional research for just 8% of the visits; that’s compared to 15% without it. AI overviews are just eating traffic continuously, but now they have become the search landscape, and we have no choice but to navigate that. Today, the SERPs are more for awareness and less about traffic. And if you’re not ranking for AI overview, you will not be seen.
When the page that’s cited for AI search response is usually displayed at the top of SERP, you’re “ranking” for AI overview. The rightly cited URL that’s visible for searchers for the desktop search without the interaction.
The major URLs that are visible if the users click “Show more” and all the cited sources that become visible if someone can expand the complete list of clicks to “Show all.”
These rankings are not the final ones. Do not think of them for traditional search rankings.
Google’s AI overviews are non-deterministic, thus meaning that they can change for each refresh, and so do use the URLs as they cite.
Here’s what we know about the ranking in AI overviews that’s based on data.
1. Targeting informational, question-based queries
I analyzed multiple SERPs and figured out AI overviews trigger for 21% of all the keywords, but the right query can trigger it much more often. For instance, AI overviews can appear for 57.9% of the question queries.
And for 46.4% of the queries for 7+ words.
The right type of question does matter. “Reason” for queries (i.e., “Why” questions) can trigger AI overviews 60% of the time, the perfect rate for categories we can study.
The “Bool” queries (for instance, the yes/no questions) along with the definition queries are higher than average for AI overviews, which rates 57.4% and 47.3% in a respective manner.
Meanwhile the 99.9% of the AI overview for keywords that’s informational.
Why does this matter? For the right understandings which can trigger the AI overviews that can help for figuring out the content types plus the topics for targeting on what that want to rank.
For instance, if the right focus topics that are informational, you can face the AI overview competition for targeting if you need to rank.
For instance, If the focus topics are transformational, you can face more AI overview competition, but you get better opportunities for citations.
Initiate it by using the Keywords Explorer for your question based keywords for your industry.
Then filter the queries for 7+ words plus informational intent and the ones that can trigger in the AI overviews. The more you can own these informational and question-based topics, the higher your chances of claiming them in AI overviews.
2. Make Sure You Rank in Traditional SERPs.
If you don’t have visibility in traditional search, you will not appear in AI overviews also. Through a large analysis of around 1.9 million AI overview citations, it was found that around 76% of rankings can be found under the top 10, with median rankings for the best-cited URLs as in position 2.
The AI overview research box of AHREFs does show organic positions distributed for the highly cited URL for AI overview, with the median position of 2 and ranging from 1 to 8. This makes complete sense as to how AI systems work. These use a system called RAG (Retrieval Augmented Generation) and pull search engine indexes towards supplementing training data.
If Google cannot surface your content for traditional search, AI overviews will not do that either. Use the organic keywords reports in Ahrefs’s Site Explorer for checking in where exactly your target pages do rank for relevant keywords.
If you’re sitting outside those top 10 rankings, focus on improvement of ranking first. Think of this as traditional SEO being a big gatekeeper for that. AI models do trust what Google trusts already. You cannot jump to the top without indexing; you need to earn your place before AI can consider citing you.
Later, you can target pages for further reference and check whether the optimizations for search do have a growing effect on your AI overview visibility.
3. Optimize for Clarity, Not Volume: Focus Towards Searcher Intent
The AI overview just does not care how long the blog is; they can care about how well the content can answer a query.
Our research can show the near-zero correlation (Spearman ~0.04) for the word count and AI citation. The main goal should be towards answering the user’s query directly and in an early-phased manner; burying the opening paragraph causes a delay in LLMs (Large Language Models) for semantic understanding for your content.
For instance, some months back I updated a piece of declining content. I thought I improved that by adding more information, thus filling in topic gaps and thus generally making that more comprehensive. But when I checked the traffic that had submerged, it was even worse. With the help of tools like Ahrefs’ Keyword Explorer and Ahrefs’ Page Inspect and Claude, I further realized I’d diluted the potency of original content by adding in too much unnecessary detail.
While I had edited the blog for better matching in the query, I added new sections for tangentially related topics. But this ended up as too much for excessive information for the straightforward question.
I did dilute the focus, thus creating a perfect mismatch, and buried important information that’s further down the page. As soon as I altered my updates, AI overview visibility towards the beginning rose again.
Based on Dan Petrovic’s latest study for Google’s Grounding (the SERP content Google uses as the source material when generating for AI overviews), he realized the intent for dilution of the real issue.
Analysis for more than 7,000 queries, he found those grounding plateaus at around 540 words, and the pages that have 2,000 words can see diminishing returns.
“Adding more content dilutes your coverage percentage without increasing what gets selected… The implication for content strategy is clear: density beats length. Focus on being the most relevant source for a query, not the longest.”
So, the basics here are do not target any specific word count and do not just randomly add more content for hedging your bets.
Let the intent and topic dictate how much content you can cite.
Here Are the Three Ways You Can Optimize for the Search Intent
1. Use Ahrefs’ Identify Intents Tool in Keyword Explorer or Other Tools Based on Your Needs
For studying the intent towards the breakdown of SERP.
For instance, around 46% of the search results for the query “AI SEO Expert” are the definitions and example guides that can help users understand the different outcomes so that it makes sense for the creation of the content that you want to rank and be in a chance for claiming an overview.
2. Use Tools Like Also Asked or Answer The Public
Use these tools for finding the questions behind those short-tail keywords and better understanding what exactly the users want to know when they can search for each query.
3. Work for Fan-Out Queries for Your Own Topics, Not Just the Keywords
Research shows that you can improve AI overviews by doing the “fan-out query” for SERPs.
What exactly is a fan-out query? Well, according to the official documentation of Google, when a user searches for something in Search and AI gets triggered, the system does a “Query fan-out” thus breaking down a single query to multiple sub-related queries.
For a user prompting as in “What will happen when you swap regular flour with wheat flour with honey drizzle?” the AI can generate those fan-out queries like “best flour for honey drizzle,” “how does wheat flour affect cake density,” and “baking with wheat flour tips.”
This allows for Gemini, the model that powers AI overviews, to explore different areas of the topic, plus retrieve passages from multiple sources for construction of well-adjusted and contextual answers.
According to the alumni of AHREFs for Joshua Hardwick, who actually researched the complete fan-out query behavior for AI overviews along with SurferSEO, pages that can rank for these fan-out queries are 161% more likely for the primary search team.
Josh also found that the Spearman correlation between ranking for the fan-out queries plus being cited for an AI overview is 0.77; that means super strong.
This is clear now. Ranking for AI overviews is all about building deep topical authority.
Do not overfocus on individual keywords plus isolated citations. Write the content that covers multiple aspects for your target topic to improve your chances of showing up for a fan-out query.
This is pretty clear: ranking for AI overviews for building deep topical authority.
Do not focus on isolated citations or individual keywords. Write the content that covers the multiple aspects for your targeted topic towards improving your chances of showing up for fan-out queries for SERPs.
A perfect way for doing this is for the optimization of fan-out queries in a direct manner, which GoFishDigital and Dan Hinkley have done.
Using the Screaming Frog and Gemini API for extraction of Google’s fan-out queries. Dan created an important Python script that you can just copy & paste to:
- Crawl certain pages for your site (i.e., the /blog/ subfolder)
- Using Screaming Frog for extraction of H1 from each page.
- Send those H1s towards Gemini.
- Extracting both the response of the AI overview plus a list of fan-out queries for each H1.
Here’s another, even easier way to extract AI overviews for fan-outs. Ofora is the exact free tool for AI experts for iPullRank founder Mike King. This is modeled on Google’s official retrieval patents and replicates Gemini’s fan-out query process that involves the use of Gemini itself for generating those queries.
1. What Happens When AI Overviews Can Currently be Associated With Your Brand
Add the brand to Brand Radar and head to the topics report and check which themes the AI overviews can currently associate with your brand.
Then spot those gaps in between, as in where you want to be vs. where the AI perceives it.
2. Checking for Topic + Entity Gaps for AI Overviews
Here’s an effective workflow done by Rodrigo Stockebrand, the Global Head of SEO for Entain and the former head of SEO for NASA.
He exported non-brand entities and AI overview keywords from Ahrefs that can report to the site explorer.
Then he uses AI Studio/Poe for the creation of network graphs that show the entity clusters that do appear in AI clusters and which don’t.
3. Build Content Clusters
Parent Topic in Ahrefs’ Keyword Explorer shows all the keywords that you can easily target with just a basic single content piece. Use that to find the cluster on your content, not just some single keywords.
Once you have created your content clusters and optimized for the existing ones, use the internal linking opportunities reports for building stronger associations, thus giving you the ready-made link recommendations.
4. Building positive brand mentions for multiple properties
AI overviews thus favor brands towards widespread recognition across videos, articles, and forums on the web.
The more properties that can reference your brand, the better chances you have of getting cited or mentioned in an AI overview. Also, according to research from AHREFs, the brand mentions are the best correlating factor for AI overview visibility.
So, how can you actually build the off-site brand mentions?
Outreach towards mentions in “best lists”
Through the marketing research of the efficacy of the best sites. I found some info regarding that. He found that the AI overviews of Google favor one of the “best” posts with more than 50% of AIO citations that can’t fall into the category.
Some points to remember:
This is practically important to understand and remember. While showing up on the “best lists” that can help for visibility, the recommendations can feel inauthentic, which can result in positive recognition, eventful ROI, and user experience.
So, be selective when you can reach out towards visibility on the lists. When that’s possible, offer writers a short film or demo (completely disclosed) so they can provide an honest review.
And instead of pitching your brand towards existing articles and bargaining for the writers to update their posts to mention your brand, try targeting authoritative sites for the niche that is not written here yet.
Ask them if they’re considering reviewing the “best X products.” You could also:
- Do the keyword research on your behalf.
- Pitch in the traffic opportunity (i.e., your competitors who are getting the X visits for the similar posts).
- Introduce the product and provide them much taste.
That way you will get a great reason for including a great value and not just asking for favors.
5. Partner with YouTube creators for product features
In the latest study done on the brand factor study done by AHREFs, the study finds out that YouTube mentions work and shows the highest correlation for AI visibility.
This makes sense as YouTube is the #1 domain that’s cited for AI overviews, as per the different study reports.
This is how you can build partnerships for YouTube to claim more in AI overview listings:
- Looking for channels producing the tutorials, educational content, and comparisons wherein, in general, the productivity hits.
- Pitch in for a specific video idea on where exactly the video idea can serve the audience, like “setting up the workflows,” if you’re a project management tool, rather than just creating for a review.
- Offering early access for features, a thorough demo, and exclusive data so they can directly integrate the product in an authentic manner.
Under a most important phase, you have to find the YouTubers that generally show up in AI overviews. If they’re mentioning your brand, then even better. That’s a great opportunity.
6. Tracking Those Videos That drive AI visibility
Using tools like brand radar for seeing which of the YouTube videos mention brands that appear for AI overviews. Under the “Cited In AI” column that shows which of the videos drive the visibility across AI assistants:
Clicking the AI assistant icon that shows you the prompt for closed doors for citations for Brand Radar’s complete AI Response report.
For instance, while analyzing the “best tool” videos, one can mention tools. I spot that LearnWire video, which turned out to be responsible for multiple mentions for AI overviews for those high-value queries such as “best SEO software” and “SEO software comparison.”
You can scale the identity to the topics, creator AI, and formats that are cited more often, then run towards the same check for competitor mentions.
From there, you can directly replicate what can work for creating similar videos and pitching in yourself towards creators also making content.
7. Get Mentioned on Highly Linked and Authoritative Pages
I analyzed multiple websites that are mentioned across millions of AI overviews and found that there is a strong correlation between mentions for AI overview visibility and high-linked pages.
Pages that can be worthy of linking for carrying in more weight, when the brand appears for authoritative sources, the overviews are significant. The key is towards earning mentions that have stronger backlink profiles, and not just some mentions.
For finding the earning mentions for pages that have strong backlink profiles and not just some mentions.
To find these kinds of pages, move to “Link Intersect: Report for Ahrefs Competitive Analysis.”
This will show up towards the high-authority pages for your industry that do mention the competitors and not you. Pitch in the guest contributions, offering expert quotes or creating resources that can offer sites for mentioning the top pages.
8. Optimise the Content for Structured Data
This is a big belief that structured data can help you to show up generally in the AI search. However, the jury is still out of context for which schema is recognised for Google’s AI model, i.e., Gemini.
LLMs do turn the numerical representations during training for the schema markup within content that can get randomised.
What’s more, generally the AI crawlers are not able to access the schema data; this is client-side rendered (for instance, through JavaScript), even though most of the sites that rely on the startup. Instead, they can tend towards looking for the RAW HTML of the page with the use of a basic snapshot, without loading everything.
Extra Tip:
Convert the JavaScript for AI crawlers to help it read that content.
If your site works on JavaScript, there’s a big workaround. Certain tools like prerender.io do show bots can show the ready-made HTML version for your page that includes schema, links and content.
But for structured data, that’s still said to have an indirect impact on AI visibility. During the Google Search Central conference, John Mueller, Senior Search Analyst, stated that structured data is important for AI search.
This is important as AI visibility works in search visibility due to the process of Retrieval Augmented Generation (RAG).
This is how the AI platforms can visit the SERP and top up their existing training knowledge information for new information; this process is called “grounding,” which was mentioned earlier.
Unlike the AI crawlers, this type of structured data is handled way better by engine bots of Google. This is also important for deeper search visibility, since that contributes to the knowledge graphs that these search engines are constructed on.
In another way of saying that, structured data here defines that you have higher chances of getting cited in SERP, which in turn puts the selection pool for AI overviews for RAG.
While I do not have considerable proof that schema markup can help you for AI overviews, for both Microsoft and Google, they have confirmed that structured data can help the AI system understand the content deeply.
In the meantime, under the given condition that doing traditional SEO best practices and implementing schema types like HowTo, Articles, and FAQPage is just a low-risk way for potentially improving the readability of the content for those AI systems.
Further, you can also use certain tools like Ahrefs’ “Structured Data” tab from the Ahrefs toolbar.
Final Thoughts
AI overviews will be here to stay, and the modern rules for ranking are being written while we talk here. Each strategy I have covered, from building topical authority to higher matching of search intent, is grounded in my analysis of thousands of AI responses, deep third-party research, plus direct guidance from Google, not some kind of speculation.
Just start with some topic-level optimization along with genuine brand-level mentions, then layer those technical improvements as you move forward.
