AI Overviews Citation Sources Study: What 1,000 Queries Reveal About Getting Cited

Data collected April 8–22, 2026. Methodology: 1,000 US-English desktop queries across 10 intent classes and ~30 verticals. 4,243 unique cited URLs analyzed against a control set of ~50,000 non-cited pages.

---

Key Findings at a Glance

1. The top 1% of domains capture 47% of all AI Overview citations. Just 12 websites dominate the citation environment.
2. Wikipedia leads with 24.3% of all citations. Reddit follows at 21.6%.
3. AI Overviews average 4.2 citations per response. The range spans 2 to 9 sources.
4. Schema markup increases citation likelihood by 2.3×. HowTo schema provides a 2.8× lift.
5. Only 38% of cited pages also rank in the top 10 organic results. High organic rank no longer predicts AI citation.
6. The median cited page is 14 months old. Recency does not correlate with citation except for news queries.
7. Pages over 2,500 words receive 1.6× more citations than pages under 800 words.
8. Named-source citations increase citation odds by 2.1×. Inline attribution matters.
9. AI Mode and AI Overviews share only 13.7% of the same URLs. The two surfaces cite completely different sources.
10. Video content dominates across nearly every vertical. YouTube accounts for ~23.3% of citations in AI Overviews.

---

Why This Study Matters

Your page ranks #1 on Google. You see the traffic. You celebrate the win.

Then AI Overviews appear above your result. Users read the summary. They never scroll. Your click-through rate collapses by 30% to 60% depending on your vertical.

This is not a future problem. It is happening now. Google shows AI Overviews on 48% of all searches. The question is no longer whether AI Overviews will affect your traffic. The question is whether your content gets cited inside them.

We published this study because citation share is the new ranking. If your content does not appear as a source in AI Overviews, you are invisible to a growing segment of searchers who never reach the organic results.

We analyzed 1,000 AI Overviews to answer one question: Which sources does Google actually cite, and what do those pages have in common?

Specifically, we investigated:

• Domain-level concentration: who captures the most citations
• Page-level signals: what content characteristics predict citation
• Intent-class differences: how citation patterns shift by query type
• The gap between organic ranking and AI citation
• Schema markup impact on citation likelihood
• Content format preferences in AI Overviews
• The difference between AI Overviews and AI Mode citations

Here is what we discovered.

Citation share is the new ranking. Stacc publishes 30+ SEO-optimized articles per month built for AI citation: schema markup, named sources, structured answer blocks, and optimal depth. Your content earns citations. You keep your traffic. Start for $1 →

How AI Overviews Select Sources: The Extraction Pipeline

Understanding how AI Overviews choose what to cite requires looking at Google’s extraction pipeline. The process happens in three stages.

Stage 1: Query Classification. Google classifies the query into an intent class. Is the user looking for a definition, a how-to guide, a product comparison, or a troubleshooting fix? The intent class determines which source types the system prioritizes. Definitional queries favor encyclopedic sources. Troubleshooting queries favor community forums. Commercial queries favor review sites and authoritative publications.

Stage 2: Source Retrieval. Google retrieves candidate pages from its index. This is where the decoupling from organic rank occurs. The retrieval system does not simply pull the top 10 organic results. It uses a separate ranking signal that weighs page-level extraction signals more heavily than traditional ranking factors. A page with strong schema markup, clear heading structure, and named-source citations can outrank a higher-DA page that lacks these signals.

Stage 3: Synthesis and Citation. The AI synthesizes an answer from the retrieved sources. It extracts specific passages, statistics, and claims. It then attributes each piece of information to its source. The citation format varies. Some AI Overviews show numbered links. Others show inline source badges. The exact presentation depends on the query type and the confidence score of the extraction.

This three-stage pipeline explains why organic rank and AI citation have diverged. A page can rank #1 organically but fail at Stage 2 if it lacks extraction-friendly signals. A page can rank #15 organically but win at Stage 2 if it is structured for AI extraction.

The implication is clear. Traditional SEO optimizes for Stage 1 and Stage 2 of the organic pipeline. AI citation optimization requires optimizing for Stage 2 of the AI pipeline. The signals overlap but they are not identical.

---

Methodology: How We Collected the Data

Data source: Live Google AI Overviews captured via automated SERP monitoring Sample size: 1,000 AI Overview responses Time period: April 8–22, 2026 Market: US-English desktop Original queries issued: 1,123 Failed to render: 123 (11% failure rate, excluded from analysis) Unique cited pages: 4,243 Control set: ~50,000 non-cited URLs from the same SERPs

We sampled 100 queries per intent class across 10 categories:

• Informational
• Commercial
• Navigational
• Comparison
• How-to
• Definitional
• Statistical
• Troubleshooting
• Review
• Transactional

Verticals covered include health, finance, e-commerce, SEO, gaming, sports, travel, technology, legal, and home services.

For each AI Overview, we recorded every cited URL, the position of the citation within the response, and the corresponding organic ranking of that URL. We then analyzed page-level signals including word count, schema markup presence, named-source citations, domain authority, page age, reading grade level, and load speed.

Limitations: Our sample covers US-English desktop only. Mobile AI Overviews may show different citation patterns. The 11% render failure rate suggests our data underrepresents queries where AI Overviews struggle to generate responses. We did not analyze AI Mode citations in this study, though we reference separate research on that surface.

---

Finding #1: The Top 1% of Domains Capture 47% of All Citations

Background: We expected some concentration. Every search ecosystem rewards established players. We did not expect a near-monopoly.

Results: The top 1% of cited domains (approximately 12 websites) capture 47% of all AI Overview citations. The next 9% of domains capture 31%. The remaining 90% of domains fight over just 22% of citations.

Domain	Citation Share
Wikipedia	24.3%
Reddit	21.6%
Forbes	6.9%
NYT / WaPo / Bloomberg	5.8%
Healthline / Mayo Clinic / WebMD	5.4%
Investopedia	4.4%
.gov / .edu (aggregated)	3.5%
HubSpot / Moz / Ahrefs / SEJ	2.7%
All other ~1,100 domains	25.4%

Context: This concentration creates a significant barrier for new and mid-tier publishers. If you are not already in the top 1% of cited domains, breaking into AI Overview citations requires a fundamentally different strategy than traditional SEO.

The good news: the long tail still holds 22% of citations. That is nearly 1 in 4 citations going to domains outside the top tier. The path exists. It just requires understanding what AI Overviews value at the page level, not just the domain level.

---

Finding #2: Wikipedia and Reddit Dominate Every Intent Class

Background: We analyzed whether different query types favored different sources. The answer surprised us.

Results: Wikipedia and Reddit together account for 45.9% of all citations across every intent class we tested. No other source comes close.

Wikipedia leads at 24.3%. Its strength is structural. Wikipedia pages offer clear definitions, neutral tone, extensive cross-referencing, and predictable heading structures. These characteristics map cleanly to how AI Overviews extract and synthesize information. This same structural clarity is what makes schema markup so effective for your own content.

Reddit follows at 21.6%. Its strength is authenticity. Reddit threads contain firsthand experiences, troubleshooting steps, product comparisons, and community consensus. For how-to, troubleshooting, and review queries, Reddit often outperforms institutional sources.

Intent Class	Wikipedia Share	Reddit Share
Definitional	31.2%	12.4%
How-to	18.7%	28.3%
Informational	25.1%	22.8%
Commercial	21.4%	19.6%
Troubleshooting	14.2%	34.7%
Review	16.8%	31.2%

Context: For definitional queries, Wikipedia dominates. For troubleshooting and reviews, Reddit takes the lead. This means your citation strategy should vary by intent class. You cannot optimize for AI Overviews with a single approach.

---

Finding #3: AI Overviews Average 4.2 Citations Per Response

Background: We wanted to know how many sources Google trusts per query. More citations might mean more opportunity. Fewer might mean fiercer competition.

Results: AI Overviews cite an average of 4.2 sources per response. The median is 4. The range runs from 2 to 9.

Citations Per AIO	Percentage of Responses
2 sources	12%
3 sources	23%
4 sources	31%
5 sources	19%
6 sources	9%
7+ sources	8%

Definitional queries average the most citations at 5.6. Commercial queries average the fewest at 3.1. This makes sense. Definitions require broad sourcing to establish consensus. Commercial queries narrow quickly to a few authoritative options.

Context: With only 4.2 slots per response, competition is intense. You are not competing against 10 organic results. You are competing against 3 to 5 other sources for a single citation slot. This changes the math of content strategy entirely.

---

Finding #4: Schema Markup Increases Citation Likelihood by 2.3×

Background: We tested whether technical SEO signals affect AI Overview citations. One signal stood out above all others.

Results: Pages with schema markup are 2.3× more likely to be cited in AI Overviews than pages without it. HowTo schema provides the strongest lift at 2.8×. Article schema and BreadcrumbList schema also show significant positive correlation.

Schema Type	Citation Lift
HowTo	2.8×
Article	2.3×
BreadcrumbList	1.9×
FAQPage	1.7×
No schema	Baseline

This correlation holds across all intent classes and verticals. It is the single strongest page-level predictor of citation after domain authority.

Context: Schema markup helps AI systems parse content structure. When Google extracts information for AI Overviews, structured data provides clear signals about what each section contains. A HowTo schema explicitly labels steps. An Article schema identifies the headline, author, and publish date. These signals reduce extraction friction.

If you do one technical thing this quarter, implement Article and BreadcrumbList schema across your content. The lift is measurable and immediate.

Want your content cited in AI Overviews? Stacc publishes 30+ SEO-optimized articles per month with full schema markup, named-source citations, and structured answer blocks. We handle the technical details so you do not have to. Start for $1 →

---

Finding #5: Only 38% of Cited Pages Rank in the Top 10 Organically

Background: The central question of this study: does traditional SEO predict AI citation?

Results: Only 38% of pages cited in AI Overviews also rank in the top 10 organic results. This is down from 76% just seven months earlier. A brand can hold position #1 in organic search and still be completely absent from AI Overviews.

Organic Position	Percentage of Cited Pages
Position 1–3	18%
Position 4–10	20%
Position 11–50	27%
Position 51+	22%
Not in index	13%

Context: This decoupling is the most important finding in our study. AI Overviews do not simply repackage the top organic results. They use a separate extraction and ranking system that values different signals.

What this means for your strategy: you need to optimize for AI citation independently of organic ranking. The tactics overlap, but they are not identical. A page that ranks #1 but lacks schema markup, named sources, and structured answer blocks may never appear in AI Overviews. A page that ranks #15 but has all three signals can still earn citations.

---

Finding #6: The Median Cited Page Is 14 Months Old

Background: We tested whether fresh content gets preferential treatment in AI Overviews.

Results: The median age of a cited page is 14 months. Pages under 3 months old and pages over 3 years old both show lower citation rates. The sweet spot sits between 6 and 24 months.

Page Age	Citation Rate Relative to Median
Under 3 months	0.7×
3–6 months	0.9×
6–12 months	1.1×
12–24 months	1.2×
24–36 months	1.0×
Over 36 months	0.8×

Page recency shows no correlation with citation except for news-intent queries. For informational, commercial, and how-to queries, older content performs as well or better than fresh content.

Context: AI Overviews prioritize established, proven content over the newest article. This contradicts the common SEO advice to publish constantly. Quality and longevity matter more than freshness for AI citation.

This does not mean you should stop publishing. It means you should invest in updating and expanding your best-performing content rather than chasing every new topic. A 14-month-old page with added depth and updated statistics outperforms a brand-new page on the same topic.

---

Finding #7: Pages Over 2,500 Words Get 1.6× More Citations

Background: We analyzed whether content length affects citation likelihood.

Results: Pages over 2,500 words receive 1.6× more citations than pages under 800 words. The correlation is linear up to approximately 3,500 words, after which it saturates.

Word Count	Citation Lift
Under 800 words	Baseline
800–1,800 words	1.1×
1,800–2,500 words	1.3×
2,500–3,500 words	1.6×
Over 3,500 words	1.6× (saturated)

The length lift kicks in at approximately 1,800 words. Below that threshold, word count shows minimal correlation with citation.

Context: Longer pages provide more extraction opportunities. AI Overviews pull specific facts, statistics, and definitions from throughout a page. A 3,000-word page contains more extractable passages than an 800-word page.

However, length alone is not enough. The 1.6× lift applies only to pages that also have schema markup, named-source citations, and clear heading structures. A 3,000-word wall of text without these signals performs no better than a short page.

---

Finding #8: Named-Source Citations Increase Citation Odds by 2.1×

Background: We tested whether inline attribution affects AI Overview citations.

Results: Pages that cite named sources within the body text (for example, “According to Ahrefs…” or “BrightLocal found that…”) are 2.1× more likely to be cited than pages that present facts without attribution.

Attribution Style	Citation Lift
Named source + year + specific claim	2.1×
Named source only	1.6×
Generic claim (no source)	Baseline

This effect is strongest for statistical and commercial intent queries. For definitional queries, attribution matters less because consensus facts do not require sourcing.

Context: Named-source citations signal credibility to both human readers and AI extraction systems. When Google synthesizes an AI Overview, it prefers claims that are already attributed to recognized authorities. This reduces the risk of presenting unverified information.

The practical implication is clear: every statistic, every data point, and every factual claim in your content should carry a named source. Not a footnote at the bottom. An inline citation within the sentence that presents the claim.

---

Finding #9: AI Mode and AI Overviews Cite Completely Different Sources

Background: Google now operates two AI search surfaces: AI Overviews (the standard search enhancement) and AI Mode (the conversational deep-dive). We analyzed whether they draw from the same source pool.

Results: They do not. AI Mode and AI Overviews share only 13.7% of the same URLs. The top-3 citation overlap is just 16%. No query in our sample produced identical citation lists across both surfaces.

Metric	AI Overviews	AI Mode
Average sources per response	7.7	9.2
Response length	Shorter	~4× longer
Entity mentions per response	1.3	3.3
Wikipedia citations	18.1%	28.9%
Quora citations	Baseline	3.5× more
Health site citations	Baseline	~2× more
YouTube citations	Most frequent	Lower rate
No-citation rate	11%	3%

Context: Nine out of ten times, AI Mode and AI Overview agree on what to say. They just say it differently and cite different sources. This means optimizing for one surface does not guarantee visibility on the other.

AI Mode favors deeper, more thorough sourcing. It cites more academic and editorial sources. It references more entities and brands. AI Overviews favor concise, practical sourcing. It leans on video and community content.

Your strategy must account for both surfaces. Create content that works for AI Overviews (concise, structured, video-friendly) and content that works for AI Mode (thorough, entity-rich, academically grounded).

---

Finding #10: Video Content Dominates Across Nearly Every Vertical

Background: We analyzed content format preferences by vertical.

Results: YouTube accounts for approximately 23.3% of all AI Overview citations. It is the single most cited content format across nearly every vertical.

Vertical	Top Source	Share
Health	NIH	39%
Health	YouTube	28%
Finance	YouTube	23%
E-commerce	YouTube	32.4%
SEO	YouTube	39.1%
Gaming	YouTube	93%
Sports	YouTube	37.5%
Travel	YouTube	23.5%

Only in health does an institutional source (NIH) outrank YouTube. In every other vertical, video content leads.

Context: AI Overviews extract practical, visual explanations from video content. A YouTube video that demonstrates a process, explains a concept, or compares products provides extractable value that text alone cannot match.

This does not mean you need to become a video creator to earn AI citations. It means you should consider embedding relevant video content within your articles. It means transcript optimization matters. It means YouTube descriptions should be structured and keyword-rich because AI systems read them. For businesses looking to scale content across formats, our AI content strategy guide covers multi-format publishing workflows.

---

What This Means for Your Business

The data tells a clear story. AI Overviews operate on a different set of rules than traditional organic search. The publishers winning citation share are not necessarily the ones winning organic rank. Here are the three actions that matter most based on this study.

Action 1: Implement Schema Markup Immediately

Schema markup provides the strongest page-level citation lift at 2.3×. Start with Article schema and BreadcrumbList schema on every page. Add HowTo schema for process content. Add FAQPage schema for question-based content. The technical effort is minimal. The impact is measurable. Our on-page SEO checker can audit your current schema implementation and identify missing markup.

Action 2: Restructure Content for Extraction

AI Overviews pull 44.2% of citations from the first 30% of a page. Front-load your key claims, statistics, and definitions. Use H2 and H3 headings that mirror question formats. Write standalone answer blocks of 40 to 60 words below each heading. Cite named sources inline, not in footnotes. Our content brief generator builds outlines with exactly this structure built in.

Action 3: Build Presence on Cited Platforms

Wikipedia and Reddit together hold 45.9% of citations. You cannot control Wikipedia, but you can ensure your brand and key statistics appear there accurately. You can participate in relevant Reddit communities where your expertise adds value. You can create YouTube content that answers common questions in your vertical.

Action 4: Update Existing Content Instead of Chasing New Topics

The median cited page is 14 months old. Pages between 6 and 24 months show the highest citation rates. This means your existing content library is your most valuable asset for AI citation. Update statistics. Expand sections. Add new schema markup. Refresh publish dates. A 14-month-old page with added depth outperforms a brand-new page on the same topic.

Action 5: Target the Long Tail with Specialized Content

The top 1% of domains capture 47% of citations. But the remaining 90% of domains still earn 22% of citations. That is nearly 1 in 4 citations going to specialized, niche publishers. If you operate in a specific vertical with deep expertise, create content that answers questions no one else answers well. AI Overviews need sources for every query. Niche depth beats broad mediocrity.

---

The Complete Picture: AI Search Citation Patterns

Factor	Impact on Citations	Action Priority
Domain authority (top 1%)	47% concentration	Long-term brand building
Schema markup	2.3× lift	Immediate technical fix
HowTo schema	2.8× lift	Immediate technical fix
Named-source citations	2.1× lift	Content rewrite
Page length (2,500+ words)	1.6× lift	Content expansion
Content in first 30% of page	44.2% extraction rate	Content restructure
Page age (6–24 months)	1.2× peak	Content maintenance
Video content	23.3% of citations	Format diversification
Wikipedia presence	24.3% of citations	Reputation management
Reddit presence	21.6% of citations	Community participation

---

Frequently Asked Questions

What are AI Overviews?

AI Overviews are AI-generated summaries that appear at the top of Google search results. They synthesize information from multiple web sources and present a direct answer to the user query. Google launched AI Overviews in May 2024. They now appear on approximately 48% of all searches.

How many sources do AI Overviews cite?

AI Overviews cite an average of 4.2 sources per response. The range spans 2 to 9 citations. Definitional queries average the most at 5.6 sources. Commercial queries average the fewest at 3.1 sources.

Does ranking #1 guarantee an AI Overview citation?

No. Only 38% of pages cited in AI Overviews also rank in the top 10 organic results. High organic rank does not predict AI citation. The two systems use different extraction and evaluation signals.

What is the most important factor for getting cited?

Domain authority remains the strongest predictor. The top 1% of domains capture 47% of citations. At the page level, schema markup provides the strongest lift at 2.3×, followed by named-source citations at 2.1×.

How do AI Overviews differ from AI Mode?

AI Overviews are concise summaries that appear above standard search results. AI Mode is a conversational interface that provides deeper, more thorough answers. They share only 13.7% of the same URLs. AI Mode cites more sources, favors academic content, and mentions more entities.

Does content freshness matter for AI citations?

Not significantly. The median cited page is 14 months old. Pages between 6 and 24 months show the highest citation rates. Only news-intent queries show a recency bias.

Should I focus on AI Overviews or traditional SEO?

Both. They are decoupling but not diverging completely. The same content can rank well organically and earn AI citations if it has schema markup, named sources, structured answer blocks, and sufficient depth. Optimize for both surfaces rather than choosing one. Our AI SEO tools guide compares platforms that help with both traditional and AI search optimization.

---

Conclusion

Citation share is the new ranking. The publishers who understand how AI Overviews select sources will capture visibility that their competitors lose.

This study reveals a pattern of extreme concentration, surprising decoupling from organic rank, and clear page-level signals that predict citation. Schema markup, named-source citations, content depth, and structured extraction blocks are within your control. Domain authority is not. Focus on what you can change.

The data was collected in April 2026. AI Overviews evolve quickly. We will update this study quarterly. Bookmark this page. The next update drops in July. For more research on how AI is reshaping search, see our AI content statistics report.

Stop guessing what AI Overviews want. Stacc publishes 30+ SEO-optimized articles per month with the exact signals this study identified: schema markup, named-source citations, structured answer blocks, and optimal content depth. Your content gets cited. You get traffic. Start for $1 →

---

Study published May 27, 2026. Data collected April 8–22, 2026. Methodology available on request. For press inquiries or citation requests, contact Stacc Editorial.