How to Rank in AI Overviews: 7 Tactics Backed by Real Data
TL;DR
- Google AI Overviews now reduce clicks to the #1 organic result by 58%, according to Ahrefs’ February 2026 study, making citation inside the overview itself the new visibility battleground.
- Pages that rank for multiple “fan-out queries” (the sub-searches Gemini runs behind the scenes) are 161% more likely to get cited in AI Overviews than pages ranking only for the primary keyword, per SurferSEO and Joshua Hardwick’s research.
- Google’s grounding system plateaus at roughly 540 words of extracted content per page, so density of relevant information beats article length every time, based on Dan Petrovic’s analysis of 7,000+ queries.
- Brand mentions across third-party sites (especially YouTube) show the strongest correlation with AI Overview visibility, according to Ahrefs’ 75K-brand study.
- Traditional organic rankings still matter: 76% of AI Overview citations come from pages already in the top 10 organic results.
The click you used to own is gone
I spent the first half of 2025 watching a slow-motion car crash. One of my best-performing blog posts, a piece that sat comfortably at position 1 for an 8,000-search-per-month keyword, started losing traffic despite not dropping a single organic rank. The culprit was a blue box at the top of the SERP that answered the query before anyone could scroll.
That blue box is Google’s AI Overview. And the numbers behind it are brutal. Ahrefs ran the same click-impact study twice: once in April 2025, when they found AI Overviews reduced clicks by 34.5%, and again in February 2026, when the number had ballooned to 58%. Pew Research Center confirmed the pattern from a different angle, reporting that users click a traditional search result in only 8% of visits where an AI summary appears, compared to 15% without one.
So the game has changed. Ranking #1 organically is still valuable (more on that in a minute), but the real prize is getting your page cited inside the AI Overview. This article covers the seven tactics I’ve found that actually move the needle, based on data from over 200 million analyzed keywords across 11 published studies. No generic “write great content” advice. You’ll walk away with specific, testable actions.
How does Google decide what to cite in an AI Overview?
Before you can optimize for AI Overviews, you need a mental model of how Gemini (the LLM powering AI Overviews) picks its sources. Think of AI Overviews like a research assistant with a very specific workflow.
Query fan-out is the process where Gemini takes your original search query and breaks it into multiple related sub-queries. Google confirmed this mechanism in their official AI features documentation. If someone searches “best project management tool for remote teams,” Gemini might internally run searches like “project management tools with async features,” “remote team collaboration software pricing,” and “Asana vs Monday vs Notion for distributed teams.” It then pulls passages from top-ranking pages across all of those sub-searches.
Grounding is the second piece. Gemini doesn’t just rely on its training data. It performs retrieval-augmented generation (RAG), grabbing fresh content from the search index to “ground” its answer in real, current sources. But here’s the part most articles miss: there’s a ceiling on how much content Gemini actually extracts from any single page. Dan Petrovic at Dejan AI analyzed over 7,000 queries and found that grounding plateaus at roughly 540 words of extracted text per page. Pages over 2,000 words see diminishing returns because the extra content dilutes the percentage of relevant material Gemini can select.
“Adding more content dilutes your coverage percentage without increasing what gets selected. The implication for content strategy is clear: density beats length.”
— Dan Petrovic, Founder, Dejan AI (Source)
That finding reframes everything. The question isn’t “how long should my article be?” It’s “how much of my page is directly relevant to what Gemini is looking for?”
Why your organic ranking still determines your AI Overview chances
I know what you’re thinking. If AI Overviews are stealing clicks from organic results, maybe organic rankings don’t matter anymore. Wrong.
Ahrefs analyzed 1.9 million AI Overview citations and found that 76% of cited URLs also rank in the top 10 organic results. The median organic position for the top-cited URL? Position 2. AI Overviews aren’t pulling from page 4 of Google. They’re pulling from the same pages that already dominate traditional search.
This makes sense when you understand RAG. Gemini’s grounding process starts with Google’s own search index. If your page doesn’t surface in traditional search, it’s not in the candidate pool for AI Overview citations. Period.
Here’s the practical takeaway: AI Overview optimization isn’t a replacement for SEO. It’s a layer on top of it. If you’re sitting outside the top 10 for your target keywords, fix that first. The fancy AI-specific tactics I’m about to cover won’t help a page that Google doesn’t already trust.
| Priority Level | Your Current Organic Position | Where to Focus |
|---|---|---|
| Fix this first | Not in top 20 | Core SEO: content quality, backlinks, technical health |
| Ready for AI optimization | Positions 5-10 | Fan-out query coverage, content density, brand mentions |
| Protect and expand | Positions 1-4 | Topic authority building, YouTube partnerships, entity optimization |
Tactic 1: Map and target fan-out queries (the single biggest lever)
This is the tactic that moved the needle most for me, and the data backs it up.
Joshua Hardwick, formerly of Ahrefs, partnered with SurferSEO to study how fan-out queries affect AI Overview citations. The headline finding: pages that rank for both the primary keyword and its fan-out queries are 161% more likely to be cited in the resulting AI Overview. The Spearman correlation between ranking for fan-out queries and earning an AI citation was 0.77, which is categorized as “very strong.”
And here’s something even more surprising from Hardwick’s data. Pages that ranked only for fan-out queries (but not the main keyword) were 49% more likely to get cited than pages ranking only for the main query itself. Let that sink in. Owning the subtopics can matter more than owning the head term.
So how do you actually find fan-out queries? Three approaches, ranked from easiest to most thorough:
- Use Mike King’s free Qforia tool. iPullRank founder Mike King built Qforia, a free tool modeled on one of Google’s own retrieval patents. Enter your target keyword and a free Gemini API key, and it generates the fan-out queries Gemini would likely produce.
- Run “People Also Ask” and “Related Searches” through intent analysis. These aren’t perfect proxies for fan-out queries, but they’re directionally similar. Tools like AlsoAsked and AnswerThePublic map the question tree behind any keyword.
- Use the Screaming Frog + Gemini API method. Dan Hinkley at GoFish Digital published a Python script that crawls your blog pages, extracts H1s, sends them to Gemini, and returns fan-out queries for each. More technical, but the output is gold.
Once you have your fan-out queries, the play is straightforward: make sure your content addresses each sub-question with a direct, clear answer. You don’t need separate pages for each fan-out. Often, a well-structured section or FAQ item on an existing page is enough.
Pro Tip: Don’t just sprinkle fan-out keywords into your text. Answer each sub-question in its own dedicated paragraph or H3 section so Gemini can extract a clean passage. The grounding system pulls discrete chunks, not scattered mentions.
Tactic 2: Increase your content density (and stop padding word count)
I used to believe longer content ranked better. We all did. And for traditional organic rankings, there was some historical truth to the correlation (though it was always more about topical completeness than word count itself).
For AI Overviews, length can actually hurt you. Remember Dan Petrovic’s finding: Gemini’s grounding extraction plateaus at about 540 words per source. That means a 3,000-word page where 540 words are directly relevant has an 18% relevance density. A 1,200-word page with the same 540 relevant words has a 45% density. Which one do you think Gemini prefers?
Ahrefs’ own team learned this the hard way. They updated a declining blog post by adding more sections and related subtopics, thinking comprehensiveness would help. Traffic dropped further. When they used Claude to diagnose the issue, the analysis flagged intent dilution and scope creep as the root cause. Reversing the additions brought AI Overview visibility back.
Here’s my framework for thinking about content density:
The “one query, one answer” test: For every section of your page, ask: “Does this directly serve the primary question this page exists to answer?” If the answer is no, it belongs on a different page or it doesn’t belong at all. Tangentially related information that you think “rounds out” the topic is exactly the kind of padding that dilutes grounding density.
This doesn’t mean writing thin content. It means writing dense content. Every paragraph earns its place by saying something the reader (and Gemini) needs to know about this specific query.
Tactic 3: Build brand mentions everywhere (especially YouTube)
Here’s where things get interesting. And where most “how to rank in AI Overviews” articles get vague.
Ahrefs studied 75,000 brands and found that branded web mentions across third-party sites showed the strongest correlation (0.664) with a brand being mentioned in AI Overviews. Not backlinks. Not domain rating. Mentions.
Their follow-up study went further. YouTube mentions specifically showed the strongest individual correlation with AI Overview visibility at 0.740. YouTube is also the single most-cited domain in AI Overviews according to Ahrefs’ Brand Radar data, outpacing every other website by a wide margin.
Why does this matter? Because Gemini doesn’t just read web pages. It reads YouTube transcripts. It reads Reddit threads. It reads forums and review sites. When your brand appears consistently across these properties, Gemini treats it as a stronger “entity” with higher authority on the topic.
What does this look like in practice? Forget cold-emailing bloggers asking for a link. Instead:
- Partner with YouTube creators who already show up in AI Overviews. Don’t pitch a generic “review my product” request. Propose a specific video concept that serves their audience and naturally features your brand.
- Get into “best of” lists. Glen Allsopp at Ahrefs found that nearly 50% of AI Overview citations come from “best” list-style posts. Instead of asking to be added to existing lists, pitch authoritative sites in your niche that haven’t written one yet. Do the keyword research for them. Show them the traffic opportunity.
- Participate in communities where AI crawlers look. Reddit, Quora, and niche forums are crawled by AI systems. Genuine, helpful participation (not spam) creates brand mentions that feed into Gemini’s understanding of your brand.
Tactic 4: Understand the volatility before you panic
One thing that took me too long to internalize: AI Overview citations are not like organic rankings. They’re wildly unstable.
Ahrefs found that 45.5% of citations change when AI Overviews update, and those updates happen roughly every two days. Search Atlas reported 40-60% of cited domains change every 30 days, with Google AI Overviews exceeding 59% monthly drift.
Why does volatility matter for your strategy? Because it means you shouldn’t measure AI Overview performance the way you measure organic rankings. Checking once and declaring victory (or failure) is meaningless. You need trend data over weeks, not snapshots.
It also means the barrier to entry is lower than you think. Unlike organic rankings, where the top 3 positions can feel locked in for months, AI Overviews reshuffle constantly. A page that wasn’t cited yesterday might get cited tomorrow if you improve its content density or earn a new brand mention. The door is always cracking open.
Watch Out: Don’t obsess over individual AI Overview citation checks. The non-deterministic nature of Gemini means your page might appear in one query refresh and vanish in the next. Track citation trends over 2-4 week windows using tools like Ahrefs Brand Radar or Semrush’s Organic Rankings AI Overview filter.
Tactic 5: Match intent with surgical precision
Semrush’s 2025 study of 10 million keywords found that AI Overviews appear for about 16% of all queries, with that number peaking at 24.61% in July 2025 before settling back. But here’s the nuance everyone misses: the intent mix is shifting fast.
In October 2024, 89% of keywords triggering AI Overviews were informational. By October 2025, that number had dropped to just 57%. Commercial and transactional queries now make up nearly half of AI Overview triggers. If you’re only optimizing informational content for AI Overviews, you’re leaving almost half the opportunity on the table.
And within informational queries, “why” questions trigger AI Overviews 59.8% of the time (the highest of any query type Ahrefs measured across 146 million SERPs). Yes/no questions hit 57.4%. Definition queries land at 47.3%. Queries with 7+ words trigger AI Overviews 46.4% of the time.
The actionable move: audit your content against these trigger patterns. Are you answering “why” questions directly? Are your product and comparison pages structured to serve commercial AI Overview queries? Is your content leading with a clear, extractable answer in the first 100 words?
| Query Type | AI Overview Trigger Rate | Your Content Needs |
|---|---|---|
| ”Why” questions | 59.8% | Direct causal explanation within first 2 sentences |
| Yes/No questions | 57.4% | Clear yes or no followed by supporting context |
| Definition queries | 47.3% | Bold term + plain-English definition in same sentence |
| 7+ word queries | 46.4% | Specific, detailed answers that match the long-tail intent |
| Commercial (“best X”) | Growing rapidly | Comparison structure with clear recommendations |
Tactic 6: Use structured data (but know its real role)
Structured data is code added to your HTML that helps search engines understand the content on your page in a machine-readable format.
There’s a debate in the SEO community about whether schema markup directly helps with AI Overviews. The honest answer: probably not directly, but it matters indirectly, and that indirect path is significant.
Google Senior Search Analyst John Mueller stated at Search Central Live that structured data remains essential for AI search because it feeds the knowledge graphs that underpin search results. Since Gemini pulls its grounding sources from those same search results (via RAG), anything that improves your traditional search visibility also improves your candidacy for AI Overview citations.
The practical minimum: implement Article, FAQPage, and HowTo schema on relevant pages. Use Google’s Rich Results Test to validate your markup. And if your site renders schema via JavaScript, know that many AI crawlers can’t execute JS. Consider server-side rendering or a prerendering solution.
Don’t let schema become a rabbit hole, though. I’ve seen teams spend weeks perfecting their structured data while ignoring content density and fan-out query coverage. Those two move the needle orders of magnitude more.
Tactic 7: Build topical authority through content clusters (not isolated posts)
Everything I’ve covered so far points to one overarching principle: AI Overviews reward topical authority, not keyword-level optimization.
Fan-out queries test whether you own the topic broadly. Brand mentions signal that others consider you an authority. Content density shows you can address the core question without wandering off-topic. All of these are facets of the same idea: Gemini trusts sources that demonstrate deep, consistent expertise on a subject.
The most effective structural move I’ve made is shifting from isolated blog posts to interconnected content clusters. One pillar page covering the core topic, supported by 5-8 detailed pages covering specific subtopics, all internally linked. When Gemini runs its fan-out queries, different pages in the cluster can surface for different sub-searches, and each one reinforces your authority on the parent topic.
This isn’t a new concept in SEO. But it’s become more directly rewarded now that we can see the fan-out mechanism at work. Build the cluster, interlink it tightly, and make sure each page in the cluster passes the density test for its specific subtopic.
Frequently Asked Questions About Ranking in AI Overviews
Can you pay to appear in Google AI Overviews?
No. Google AI Overviews are entirely organic. You can’t buy placement inside the AI-generated summary. Paid ads sometimes appear above or alongside AI Overviews, but the citation sources within the overview itself are selected algorithmically based on content quality, relevance, and authority signals.
How long does it take to start appearing in AI Overviews?
There’s no fixed timeline. Because AI Overview citations are non-deterministic and change roughly every two days, a page can begin appearing within days of being indexed if it matches the right query and intent. In practice, building the topical authority and brand mentions needed for consistent citation takes 2-6 months of focused effort.
Do AI Overviews appear for every Google search?
No. According to Semrush’s 2025 study of 10 million keywords, AI Overviews appear for roughly 16% of all US desktop queries. They trigger most often for informational and long-tail queries, but Google has been expanding them into commercial and transactional searches throughout 2025 and into 2026.
Is optimizing for AI Overviews different from optimizing for ChatGPT or Perplexity?
Yes, with significant overlap. All AI search engines favor well-structured, authoritative content. But Google AI Overviews specifically use the traditional Google search index for grounding, which means your organic ranking directly affects your AI Overview chances. ChatGPT and Perplexity have their own crawling and citation systems. Brand mentions and content quality help across all platforms, but the fan-out query mechanism is specific to Google’s Gemini.
Should I block AI crawlers from my site?
For most websites, no. Blocking AI crawlers removes you from the candidate pool for citations across AI Overviews, ChatGPT, Perplexity, and other AI search tools. Unless you have a very specific reason (like protecting paywalled content), the visibility benefit of being cited far outweighs the click-rate cost in most cases.
What to do this week
If you’re going to take one thing from this article and act on it today, make it fan-out query mapping. Pick your five most important keywords, run them through Qforia, and compare the sub-queries against what your existing content actually covers. I’d bet you’ll find gaps. Fill them with dense, directly relevant answers, and you’ll be working the single strongest lever the data shows us.
The bigger picture? AI Overviews aren’t a separate channel to optimize for. They’re the visible output of a system that rewards exactly what good SEO has always rewarded: deep expertise, clear answers, and genuine authority. The difference is that now you can see the mechanism (fan-out queries, grounding extraction, brand-entity recognition) and optimize for it specifically.
If you’d rather have a team handle the research, content engineering, and ongoing monitoring, LoudScale builds AI search visibility strategies grounded in this exact data.
The clicks are shifting. The question is whether they’re shifting toward you or away from you.