62% of AI Overview Citations Come From Outside the Top 10

In July 2025, roughly 76% of pages cited in Google AI Overviews also ranked in the top 10 organic results. Rankings were a reasonable proxy for AI visibility. If you ranked well, you’d probably get cited.

Eight months later, that number sits at 38%.

I’m Gabriel Bertolo, founder of Radiant Elephant, a boutique SEO and GEO agency in Northampton, Massachusetts. We’ve been tracking this shift across every client engagement since it started accelerating, and the speed of the change is what makes it significant. This isn’t a gradual drift. It’s a structural break in how Google’s AI selects sources, and it has direct implications for how we build content strategies for clients across Western Massachusetts and nationally.

I covered this finding as part of the research review ranking all 15 GEO tactics by evidence quality. This article goes deeper on the data, what’s driving the shift, and what it means for your content strategy.

The data on how fast this changed

Ahrefs’ March 2026 study examined 863,000 keyword SERPs and 4 million AI Overview URLs. Only 37.9% of cited URLs ranked in the first 10 organic result blocks. Down from approximately 76% in July 2025. That’s a drop of nearly 40 percentage points in eight months.

The remaining citations split almost evenly: 31.2% from positions 11-100 and 31.0% from pages ranked beyond position 100.

Surfer SEO’s independent analysis of 10,000 keywords found an even starker number: 67.82% of AI Overview cited pages didn’t rank in Google’s top 10 for the query or any of its fan-out queries.

BrightEdge’s 16-month tracking study tells the same story from a different angle. Overall citation-organic overlap grew from 32.3% to 54.5%, but only 16.7% of citations come specifically from top-10 results. The “sweet spot” is positions 21-100. Pages with enough authority to be indexed and trusted but that don’t hold traditional top-10 rankings.

This represents a shift from “rank #1 to win” to “be comprehensive and authoritative to win.” And it validates exactly what we saw in our GEO case study: a DR 21 site earning AI citations ahead of DR 92 competitors. The client didn’t rank #1 for most of the queries they got cited for. They ranked positions 15-60. But their content was structured for extraction, packed with data, and covered the topic comprehensively.

Why did the shift happen?

Three factors, all accelerating.

Gemini 3 upgrade (January 27, 2026). Google made Gemini 3 the global default model powering AI Overviews. The upgrade replaced approximately 42% of previously cited domains and delivers about 32% more source URLs per AI Overview response. More sources per response means the system reaches deeper into its index for every answer.

Query fan-out. When AI Overviews are triggered, Google decomposes the original query into 8-16 related sub-queries and evaluates content across all of them. A page ranking position 40 for a closely related sub-query can end up cited in an AI Overview triggered by a different primary query. This is the same fan-out mechanism that makes topical authority the strongest predictor of AI citation. Comprehensive topic coverage means you match more sub-queries, which means more citation slots.

Expanding coverage. AI Overviews now appear on 48% of tracked queries (BrightEdge, February 2026). YMYL industries see the highest penetration: Healthcare 88%, Education 83%, B2B Technology 82%. More queries with AI Overviews means more citation slots to fill across a wider range of topics.

The CTR impact and why being cited matters more than ranking

The click-through-rate data makes the stakes concrete.

Seer Interactive studied 3,119 informational queries across 42 organizations and 25.1 million organic impressions. Organic CTR dropped from 1.76% to 0.61% for queries with AI Overviews. A 61% decline. Even queries without AI Overviews saw a 41% CTR decline (from 2.72% to 1.62%), suggesting AI’s presence is depressing click behavior across the board.

But brands cited within AI Overviews saw 35% higher organic CTR and 91% higher paid CTR than non-cited brands. Being in the AI Overview isn’t just about AI traffic. It’s about preserving the organic and paid traffic you already have. Not being cited is the real penalty.

Google AI Mode pushes this further: 93% zero-click rate (Semrush), with 60%+ of cited domains and 80% of cited URLs changing between runs.

The math here applies to every business. If you’re generating $50,000/month from organic search and AI Overviews cause a 61% CTR decline on your highest-value queries, you’re looking at potentially $30,000/month in lost revenue. Getting cited in those AI Overviews doesn’t just recover the loss. The 35% CTR boost for cited brands means you come out ahead.

How to optimize for citation without top-10 rankings

The strategy shifts from single-keyword dominance to comprehensive topic coverage and passage-level extraction.

Build for fan-out queries. Don’t just optimize for your primary keyword. Build content clusters that address the second, third, and fourth questions AI generates when expanding the original query. Surfer SEO found pages ranking for fan-out sub-queries are 161% more likely to be cited. Fifty-one percent of all AIO citations go to pages ranking for both the main query and at least one fan-out query.

Optimize at the passage level. AI Overviews extract self-contained passages, not full pages. The optimal passage length is approximately 134-167 words, with 62% of featured content landing in the 100-300 word range. Each section should answer a specific question completely enough that AI can cite it in isolation. This connects directly to answer-first content structure: the same 120-180 word, question-based H2 approach that works for overall AI citation works specifically for AI Overview extraction.

Create YouTube companion content. Among AI Overview citations from pages outside the top 100, 18.2% are YouTube URLs. YouTube provides an independent citation pathway that doesn’t require traditional organic rankings. Corrected transcripts and question-based chapter structure are the levers.

Track citations directly, not just rankings. Rankings are no longer a reliable proxy for AI visibility. Use Surfer SEO’s AI Tracker, Otterly, or Ahrefs Brand Radar to directly measure AI Overview citation rates. Track trends over 30+ day windows. A single query snapshot tells you almost nothing because of the citation drift. Multi-platform tracking matters even more now that AI Overviews are pulling from such a wide source pool.

Five implementation examples

Example 1: SaaS comparison page (position 22)

Current state: Ranks position 22 for “[Product A] vs [Product B].” Comprehensive 3,000-word article but structured as a narrative essay. Good information buried in flowing paragraphs.

Optimization: Restructure so each section is a self-contained passage answering one specific sub-query. “How much does [Product A] cost compared to [Product B]?” with a 130-word passage including specific tier-by-tier pricing and a total cost comparison for a team of 20. “Which has better customer support?” with a 110-word passage citing specific response times and G2 ratings. Each passage delivers a complete answer an AI can extract and cite independently.

Why this works at position 22: The page doesn’t need to reach position 1. It needs to have the best passage for specific sub-queries. When AI generates fan-out sub-queries about pricing, support, and features separately, each passage can earn its own citation.

Example 2: Local service page (position 14)

Current state: A plumbing company’s drain cleaning page ranks position 14 in the Northampton area. Page is 800 words of general service description with no extractable answer passages.

Optimization: Add self-contained passages for fan-out sub-queries. “How much does drain cleaning cost in Northampton?” (77 words with specific price ranges). “How long does drain cleaning take?” (72 words with specific timelines). “When should you call a plumber vs try to fix it yourself?” (90 words with specific troubleshooting steps). Each passage answers the question completely in the opening sentence.

Why this works at position 14: For local service queries, AI Overviews often cite pages outside the top 10 because the top positions are dominated by aggregator directories (Yelp, Angi, HomeAdvisor). A local business page with specific, locally-priced answer passages can outperform those directories for passage-level citation.

Example 3: Healthcare condition page (position 8)

Current state: A dermatology practice’s acne treatment page ranks position 8. Good organic traffic but not cited in AI Overviews because the content is organized as a textbook narrative without extractable passages.

Optimization: Restructure around the fan-out sub-queries AI generates. “What causes acne in adults?” (91 words with AAD statistics and practice-specific data). “How long does acne treatment take to work?” (95 words with specific timelines for each treatment type and practice-specific outcomes). “What is the best acne treatment for sensitive skin?” (95 words with specific product recommendations, concentrations, and a clinical study citation).

Why this matters for YMYL: Health content faces the highest citation bar. But the passage-level approach works in the practice’s favor. A board-certified dermatologist providing specific treatment timelines from their own clinical experience (“In our practice, the average patient achieves 70-80% clearance by week 12”) produces exactly the kind of passage AI needs to cite for health queries.

Example 4: B2B capabilities page (position 45)

Current state: A CNC machining company’s capabilities page ranks position 45 for “custom CNC machining services.” Describes capabilities in general terms. Zero AI citation potential in current form.

Optimization: Add passages targeting the technical sub-queries engineers ask. “What tolerances can CNC machining achieve?” (97 words with material-specific tolerance ranges). “How much does custom CNC machining cost per part?” (88 words with real cost ranges by complexity level). “What is the typical lead time for custom CNC parts?” (85 words with specific timelines by quantity).

Why this works at position 45: An engineer asking an AI about CNC tolerances generates fan-out queries that these passages match precisely. Most competitor pages at any ranking position provide generic capability descriptions. The passage with material-specific tolerance data (aluminum at ±0.0005″, stainless at ±0.001″, with named equipment and calibration standards) provides information gain no generic page can match.

Example 5: Financial content page (position 6)

Current state: A financial advisory firm’s Roth IRA guide ranks position 6. Strong organic performance but not getting cited in AI Overviews because the content is structured as a continuous narrative.

Optimization: Create optimized passages for each fan-out sub-query. “What is the Roth IRA contribution limit for 2026?” (88 words with exact dollar amounts and combined limit rules). “What is the Roth IRA income limit for 2026?” (95 words with specific phase-out ranges for both filing statuses). “Should you contribute to a Roth IRA or traditional IRA?” (107 words with specific income-based guidance and firm-specific data: “approximately 65% of our clients under 45 benefit from Roth contributions based on tax-projection modeling across 1,400 client cases”). “Can you withdraw from a Roth IRA before 59½?” (89 words with specific exception categories).

Why passage optimization at position 6 matters: Even a page ranking 6th can miss AI Overview citation if the content isn’t extractable. A single Roth IRA query generates 8-16 fan-out sub-queries. Each passage captures a different citation slot. The firm-specific data (“1,400 client cases”) adds information gain that generic financial sites can’t match, making each passage the preferred citation source for its specific sub-query.

The game changed. Rankings are no longer the proxy.

Rank #1 used to be the goal. Now it’s a piece of the puzzle, and not even the biggest piece. Sixty-two percent of AI Overview citations come from outside the top 10. The opportunity lives in comprehensive topic coverage, passage-level extraction, and the kind of data-rich, expert-authored content that gives AI systems something worth citing regardless of where the page ranks.

I mapped this shift alongside the full evidence base for every proven GEO tactic.