Last updated June 2026.
Every GEO practitioner eventually asks the same question: which AI engine cites what, and how many sources does each one pull per answer? The data now exists to answer it. This index consolidates the best available figures on citation volume, source-type preferences, Google-search alignment, and cross-engine overlap, so you have a single reference point rather than a dozen disconnected vendor blog posts.
This piece synthesizes published third-party research. See the attribution notes below each table for the source of every figure.
Master comparison table: citation behavior by engine
| Metric | ChatGPT | Perplexity | Gemini | Google AI Overviews |
|---|---|---|---|---|
| Avg. citations per answer | ~7.92 | ~21.87 | Not published | ~7-8 domains |
| Overlap with Google top-10 | ~8% | ~29% | ~9% | N/A (is Google) |
| Dominant cited source type | Editorial / reference | Community (Reddit) | Editorial / authority | Video (YouTube) |
| Freshness sensitivity | Moderate | High | Moderate | Moderate |
| Cross-engine domain overlap (vs. ChatGPT) | Baseline | ~11% shared | Not published | Not published |
Sources: Citation-per-answer averages: Qwairy, “Perplexity vs ChatGPT: AI Citation Study (Q3 2025),” 118,101 AI-generated answers. Google top-10 overlap: Ahrefs, “Only 12% of AI Cited URLs Rank in Google’s Top 10,” 15,000 queries, August 2025. Google AI Overviews domain average: Profound citation dataset, August 2024-June 2025. Cross-engine overlap: Profound, “Answer Engine Citation Overlap Strategy,” July 2025, 100,000 prompts. YouTube / Reddit source-type data: Surfer SEO AI Citation Report (46 million AIO citations, 2025); Profound Q2 2025 commercial-query analysis. All figures are from single-vendor studies and have not been independently peer-reviewed.
Citation volume: Perplexity cites nearly 3x more sources than ChatGPT
According to Qwairy’s Q3 2025 analysis of more than 118,000 AI-generated answers, Perplexity averaged 21.87 citations per response while ChatGPT averaged 7.92. That is a ratio of roughly 2.76 to one.
The practical implication is not that Perplexity is a better source of traffic. It is that the citation threshold differs dramatically by engine. A domain that earns a slot in a long Perplexity answer competes against 20-plus other sources in that single response. A domain that earns a slot in a ChatGPT answer is one of eight. Citation volume does not equal citation value.
Google AI Overviews typically cites a smaller, more concentrated set of domains per response. Profound’s dataset of 680 million citations (August 2024-June 2025) puts the figure at roughly seven to eight cited domains per AI Overview, comparable to ChatGPT’s average. No comparable per-answer figure has been published for Gemini at scale.
Source-type preferences: the engine-by-source breakdown
Different engines weight different source types. The table below summarizes directional findings from the best available research.
| Source type | ChatGPT | Perplexity | Google AI Overviews | Note |
|---|---|---|---|---|
| Wikipedia / reference wikis | High (top tier) | Moderate | Moderate | 5W PR synthesis, May 2026 |
| Reddit / community forums | Low-moderate | Dominant (46.7% of top-10) | Low | Profound Q2 2025, commercial queries |
| YouTube | Low | Low | Dominant (~23%) | Surfer SEO, 46M citations, 2025 |
| Brand-owned editorial | Moderate | Moderate | High | Multiple sources, directional |
| News / earned media | High | High | High | Multiple sources, directional |
The Reddit finding for Perplexity is the sharpest signal in the available data. According to Profound’s Q2 2025 study of commercial queries, Reddit appeared in 46.7% of Perplexity’s top-10 most-cited sources, making it the single most dominant source type by a wide margin. YouTube came second at roughly 13.9%.
The YouTube finding for AI Overviews comes from Surfer SEO’s 2025 AI Citation Report, which analyzed 46 million AI Overview citations. YouTube was the single most-cited domain across nearly every industry studied, accounting for approximately 23.3% of citations overall.
Both findings are from single-vendor datasets. The directional signal is clear, but treat the specific percentages as estimates rather than settled benchmarks.
Google top-10 alignment: Perplexity is the most SEO-adjacent engine
One of the most actionable findings in the available research is how differently each engine relates to traditional Google organic rankings.
An Ahrefs study of 15,000 long-tail queries (August 2025) found:
- Perplexity: ~29% of cited URLs also appear in Google’s top 10 for the same query.
- Gemini: ~9% overlap.
- ChatGPT: ~8% overlap.
- Copilot: ~8% overlap.
- Average across all four: ~12% overlap.
The gap between Perplexity and the other three engines is significant. For brands with strong organic search authority, Perplexity is the engine where traditional SEO signals translate most directly into AI citation presence. For the other three engines, SEO authority still matters, but it is a weaker predictor: roughly 92% of what those engines cite does not rank in Google’s top 10.
Cross-engine overlap: 89% of your citations do not transfer
The most important number in this index for GEO strategy is not how many citations each engine provides. It is how little the citation pools overlap.
According to Profound’s analysis of 100,000 prompts run across ChatGPT and Perplexity (July 2025), only 11% of cited domains appear in both engines. The remaining 89% of citations come from completely different sources depending on which engine a user queries.
That figure has two strategic implications.
First, earning a citation on ChatGPT does not meaningfully increase your probability of being cited on Perplexity, and vice versa. The two systems are, in terms of source preference, largely independent.
Second, a GEO programme that tracks one engine is invisible to 89% of what the other engine is doing. Multi-engine monitoring is not a premium feature; it is the minimum floor for a real citation strategy.
| Engine pair | Cited-domain overlap |
|---|---|
| ChatGPT and Perplexity | ~11% (Profound, July 2025, 100k prompts) |
| Google AI Overviews and AI Mode | ~13.7% (Ahrefs, Dec 2025, 540k query pairs) |
| AI engine citations vs. Google top-10 | ~12% average (Ahrefs, Aug 2025, 15k queries) |
The Google AI Overviews vs. AI Mode finding is particularly notable. Even within Google’s own ecosystem, two systems answering the same query cite the same URL only 13.7% of the time, according to Ahrefs’ analysis of 540,000 query pairs (December 2025). The implication: being cited in AI Overviews does not guarantee citation in AI Mode, and optimizing for one Google product does not cover the other.
Freshness sensitivity
No peer-reviewed, cross-engine freshness study with comparable methodology exists for all four engines. What the available research shows directionally:
Perplexity is described by practitioners as the most freshness-sensitive of the four, consistent with its real-time web-retrieval architecture. It actively pulls from live web pages rather than a fixed training corpus, which means recently published or updated content has a faster path to citation.
Google AI Overviews show mixed signals. Seer Interactive’s vendor study of 5,000+ URLs found that roughly 85% of AI Overview citations came from content published in 2023-2025. A larger Ahrefs study of 17 million AI citations found AI Overviews actually cited the oldest content on average compared to other AI platforms. The two findings conflict, and the Ahrefs dataset is larger and more methodologically disclosed.
ChatGPT (in its default web-search mode) retrieves live web pages for browsing-enabled queries, but its base model answers without retrieval in many contexts. Freshness sensitivity therefore depends on whether the user’s query triggers live search.
Gemini operates similarly, with real-time search grounding available but not always activated.
The practical read: if freshness is a priority signal for your category, Perplexity is the engine where a rapid-publish strategy shows results fastest.
What each engine wants: the strategy read
| Engine | Optimize for | De-prioritize |
|---|---|---|
| ChatGPT | Editorial authority; reference-quality content; third-party coverage on trusted domains | Pure SEO link building with low-authority domains |
| Perplexity | SEO-authority signals; Reddit presence and community discussion; fresh content | Video-only strategies; content absent from community platforms |
| Gemini | Editorial authority; E-E-A-T signals; structured data | Thin content without expert attribution |
| Google AI Overviews | YouTube presence; high-authority editorial; Featured Snippet-ready content | Community-first strategies with no video or editorial component |
The four engines reward meaningfully different source mixes. A Reddit-first strategy built for Perplexity will not transfer to AI Overviews. A YouTube strategy built for AI Overviews will not transfer to ChatGPT. The citation pools confirm this: 89% of citations do not cross the ChatGPT-Perplexity boundary.
Tracking citation behavior across engines
Because citation behavior differs so sharply by engine, any monitoring setup needs to cover all four. Tools that track a single engine give you a partial picture at best.
The platforms that cover multi-engine citation tracking include Profound, which built the citation-overlap analysis cited in this piece and tracks nine or more engines with deep source attribution. Peec AI covers five platforms including Google AI Overviews and AI Mode, with a large dataset of commercial queries. Otterly.AI covers six platforms with prompt-level citation tracking and a GEO audit engine. Temso covers eight engines (ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode, Grok, Microsoft Copilot, and Meta AI) from a single flat subscription, including citation monitoring that shows which source URLs each engine pulls from.
For the full comparison of GEO monitoring platforms across these dimensions, see /rankings/geo-tools. For definitions of the terms used in this piece, the /glossary covers citation share, share of model, and engine-specific retrieval concepts.
Methodology and caveats
Every figure in this index comes from a named primary source. None of it is original research by GEO Rankings.
Key caveats to keep in mind:
- The Qwairy citation-per-answer figures (21.87, 7.92) come from a single commercial vendor. They have not been independently replicated. Treat them as directional estimates.
- The Profound 11% overlap figure covers root-domain overlap, not individual URL overlap, and comes from a single vendor’s proprietary dataset.
- Source-type preferences vary by query category. Reddit’s dominance in Perplexity reflects commercial queries specifically; informational or navigational queries show different distributions.
- All figures reflect conditions in Q3 2025 through early 2026. Engine behavior changes as retrieval architectures are updated.
For our scoring criteria and quarterly update cadence, see /methodology.
Want to see where your domain sits in each engine’s citation pool? The /rankings/geo-tools ranking covers the platforms that give you engine-by-engine citation visibility, from Profound’s enterprise citation maps to Temso’s all-in-one monitoring from $29/mo.
