Multi-Engine AI Citation Overlap: Among Sources Cited 10+ Times, 98.5% Reach Two or More Answer Engines

Among sources AI answer engines cite often enough to measure, the engines largely agree. The Machine Relations Index measures citation across six AI engines — ChatGPT, Claude, Gemini, Perplexity, Google AI Mode, and Google AI Overviews. Of domains cited at least ten times in 128 days, at least 98.5% are cited by two or more engines and at least 15.4% by all six; among the 508 domains the Index grades, none is cited by a single engine. Across the whole index, at most 67.4% of cited domains are observed on only one engine, and that figure is set by the long tail: 46.2% of the index was cited exactly once, and a domain cited once is single-engine by arithmetic. Two of the six engines had collection gaps inside this window, which is why the engine-count shares above are stated as bounds rather than as measurements.

Correction, September 26, 2026. Gemini is a third engine with dated collection gaps, not a continuously collected one: 22 blind days inside the scoring window (May 11–15, then scattered single days from August 11 to September 16), carrying 442 eligible successful runs that produced no citations. The intermittent shape is why it went unnoticed while the two Google outages were found. Three arms collected without a gap — ChatGPT, Claude and Perplexity — and three are gapped: Gemini, Google AI Mode and Google AI Overviews. (An earlier form of this note, posted the same day, named only ChatGPT and Perplexity. Claude's single flagged date carried one answer run and was an artifact of a zero-citation scan with no run floor, not an outage.) Every Gemini count on this page is therefore a floor and every Gemini absence a ceiling, and no ranking that places another engine above Gemini is stated here. Rankings that place an engine below Gemini still hold, because a floor above a fixed count can only move further above it.

Correction, September 25, 2026. An earlier version of this page ranked the six engines against each other by the size of each engine's exclusive pool and by how many distinct domains each cited, and reported its engine-count shares as measurements. Every ranking on this page that places either Google surface against another engine is withdrawn. The Index's own collection recorded zero citations from Google AI Overviews on any monitored query between August 15 and September 24, 2026, covering 37 days of this release's 128-day window, and zero from Google AI Mode between September 14 and September 24, covering 7 days of it. The cause was a defect in the Index's own Google collection path, replaced on September 25, 2026: the prior collector graded a response as a success whether or not it carried content. A distinct-domain pool only accumulates, so missing collection days can remove domains from an engine's set but never add them. Google AI Mode's 7,441 and Google AI Overviews' 2,279 are floors, not measurements, and neither engine's position in any list on this page is a ranking. The same arithmetic runs one way on the agreement ladder: missing days push domains toward fewer engines, so every single-engine share on this page is an upper bound and every two-or-more and all-six share is a floor. The conditioned finding this page is about survives it — a bound in the direction of more agreement, not less. Every ChatGPT, Claude, Gemini and Perplexity figure stands unchanged; all four collected continuously through the window. The cross-engine agreement study reads the same release with the same marking.

Engine Agreement Rises With Evidence #

Release mri_score_v2.0+2026-09-21+d73e239e7c5b, window May 10 through September 21, 2026, 128 observed days, 16,146 observed answer runs, 22,377 cited domains, 113,112 domain-run citations.

On this release, 10,334 of the 22,377 cited domains — 46.2% of the index — were cited exactly once across the whole 128-day window. A domain cited once can be attributed to exactly one engine; there is no second citation available to come from a second engine. Those 10,334 domains are single-engine by arithmetic, not by retrieval behaviour, and they are 68.5% of the single-engine population. Together they carry 9.1% of all citation volume.

Condition the count on evidence and the distribution shows how much the engines share:

Source: Machine Relations Index v2, release mri_score_v2.0+2026-09-21+d73e239e7c5b. The engine count is whole-window per-engine observation in the v2 public release — which engines were observed citing the domain. The release does not publish per-engine citation counts per domain. On this release the two Google surfaces have collection gaps inside the window, so every figure in the "Cited by one engine" column is an upper bound, every figure in the "Cited by all six" column is a floor, and the mean is a floor; see the correction above.

At ten citations — still a low bar, roughly one citation every thirteen days — single-engine sources are 1.5% of what remains. At thirty they are none. The mean number of engines per domain rises from 1.64 across the whole index to 5.15.

This is why the Index conditions overlap on evidence. Counting how many distinct things a sample contains, when most of them appear once, is the singleton problem in species-richness estimation: the once-observed count dominates the tally and tracks sampling effort rather than what is out there, which is why estimators such as Chao's correct for it and why Good–Turing frequency estimation assigns singletons to unseen mass instead of treating them as findings. An AI-citation index is the same shape of data.

The per-engine exclusive pools sit in the tail. Computed over the whole index and over domains cited at least ten times:

† Floors, not measurements, and not ranked against the other engines or against each other. Both Google surfaces have collection gaps inside this window and a domain pool only accumulates, so each engine reached at least this many domains and may have reached more. See the correction of September 25, 2026 above.

Gemini and Claude are each recorded holding only two exclusive sources once a source has to have been cited ten times, and the largest remaining pool is eighteen domains. Claude's figure is exact; Gemini's blind days mean its own is not, though the direction of the finding — that conditioning on evidence collapses every exclusive pool to near nothing — is not one a collection gap can manufacture, because a gap cannot remove domains from the other engines that also cited them. Both Google surfaces read zero on the same column, which is also what a collection gap produces, so neither zero is read here as a property of the engine.

What the Conditioned Data Means for Strategy #

Engines share their heavily cited sources. Of the 508 graded domains on the September 21 release, none is cited by a single engine, and pairwise overlap between engines runs 57.3% to 90.6% among them, against 8.5% to 23.4% across the whole index. One engine sees most of what the others see, once the source is one an engine returns to. Optimizing for a single engine's preferences is a limited strategy for thin sources and small citation counts, not for the sources a brand is competing with.

Engine count and citation rate are one relationship. Domains cited by all six engines have a much higher mean citation rate than single-engine domains, but a domain's engine count cannot exceed its citation count. Being cited more is what makes multi-engine citation arithmetically available, so the gradient describes the index rather than showing that breadth causes frequency.

The evidence floor and the number of engines observed are one filter. On the September 21 release the lowest citation count among the 508 graded domains is 30, and the highest citation count of any single-engine domain anywhere in the index is 27. The two populations do not touch.

Real engine disagreement lives among heavily cited sources. Source role predicts how many engines are observed citing a domain: vendor-owned sources are observed on 2.55 engines on average against 1.54 for uncategorized ones. And on the September 21 release reddit.com is the most-cited domain in the index at 2,021 cited runs, carries an A grade, and is still absent from ChatGPT and Claude across the whole 128-day window. It is one of 105 graded domains missing two or more engines. That is the population worth studying — 105 domains, not 15,000.

Measure all six engines. A brand's own citation count is small, and at small counts one engine is a bad estimator of six.

What Independent Studies Count #

Independent studies that count each cited domain once report the same whole-population shape. SurfacedBy analyzed 127,198 citations across five engines and found 69.6% of domains appeared in only one, with 2.7% reaching all five. A separate analysis of 10,000 records across four engines found that only 4.4% of 1,843 unique domains were cited by all four, with same-query same-day overlap averaging 1.2% and a median of 0.0%. Temso AI's analysis of over two million citations found that 71% of AI sources are exclusive to a single model, and even the two most-overlapping engines (AI Overviews and Grok) agreed on only one in five sources. A DEV Community study of 412 queries found 12% shared citations across engines.

All of them count a cited domain the same way whether it was cited a thousand times or once, so they describe the long tail. Conditioned on how often a source is cited, the Index shows the engines converging.

The May–July Snapshot #

The Index's May–July 2026 window measured 15,896 domains across the same six engines. The tables below are from that window.

Source: Machine Relations Index v2, 15,896 domains measured across six AI answer engines. Data window: May–July 2026.

What Predicts Multi-Engine Citation? #

Source type is the strongest structural predictor in the MRI data. The percentage of domains reaching four or more engines varies by more than 5x across source categories:

Source: Machine Relations Index v2. Source role classifications based on the MRI taxonomy.

Community platforms and vendor-owned sources reach multiple engines at the highest rates. The 12,704 domains in the "other" category — the long tail — manage multi-engine citation at barely one-fifth the rate of editorial publications.

The domains that do reach all six engines share a pattern: Reddit, YouTube, LinkedIn, Medium, Gartner, Forbes, and G2 lead the list. These are sources that AI engines encounter repeatedly across different query types and verticals, not sources optimized for any single retrieval system.

Citation Rate by Engine Count #

In the May–July window, all-six-engine domains had a mean citation rate 70 times that of single-engine domains:

Source: Machine Relations Index v2. Citation rate = fraction of observed answer runs in which the domain was cited.

Because engine count is bounded by citation count, this gradient describes the index rather than a causal effect of breadth; see engine agreement rises with evidence.

The confidence tiers follow the same filter. Every single-engine domain in the MRI remains in the "collecting" confidence tier, meaning its citation data has not yet cleared the evidence floor for a stable rate. Among the 252 all-six-engine domains, 56% have reached a scored confidence tier (C, B, or A), with the top six earning A-tier confidence — the highest evidence grade the index assigns.

Why Engines Disagree on Sources #

Each AI answer engine uses a different retrieval pipeline, different training data, and different citation selection logic. Google AI Mode and AI Overviews both run on Google infrastructure but select different source types at measurably different rates. A Mintec analysis of 40,000 data points confirmed that AI Mode, AI Overviews, and ChatGPT cite substantially different sources for identical queries. Perplexity and ChatGPT overlap on fewer than one in five sources, even when answering identical queries.

The LIFE study measured pairwise Jaccard similarity coefficients ranging from 6.0% (ChatGPT vs. AI Overviews) to 18.9% (Perplexity vs. AI Overviews). Frase.io's analysis of citation patterns across five engines found similar divergence: each engine maintains distinct source preferences shaped by its retrieval architecture. These are not high-agreement pairs — they are the range between "almost no overlap" and "minimal overlap."

Those studies measure per-query and whole-population divergence. Among sources cited often enough to carry a published rate, the engines largely agree — no graded domain in the Index is single-engine — so the divergence matters most for thin sources and small citation counts. A source that appears only in Perplexity results reaches one retrieval context. A source cited by all six engines has crossed multiple independent selection thresholds — different retrieval models, different reranking logic, different answer generation architectures all independently deemed it citation-worthy. Research on citation behavior patterns across ChatGPT, Google AI, and other platforms shows that each engine's retrieval pipeline produces a measurably distinct citation fingerprint.

What Multi-Engine Domains Have in Common #

The 252 domains cited across all six engines are not a random sample. They cluster around several structural characteristics:

High topical breadth. These domains are not niche. The top all-six-engine sources (Reddit, YouTube, LinkedIn, Medium, Gartner, Forbes, G2, NIH, Crunchbase) span technology, business, finance, health, and general knowledge. They appear as relevant sources across multiple subject categories in the MRI taxonomy. An analysis of 366,087 citations across 12 AI models found that the publications each engine trusts are distinct — but the handful that appear everywhere share this cross-category relevance.

Established authority signals. Every domain in the all-six-engine group has been observed on at least 26 of the 71 days in the measurement window (temporal consistency above 0.37). These are not sources that spike and vanish — they are steady citation recipients over time. Search Engine Land's analysis of LLM visibility tactics confirms that consistent, authoritative content earns repeated AI engine selection — a pattern the MRI data quantifies at scale.

Content depth at scale. The long-tail "other" category produces multi-engine citation at just 0.7% — one-sixth the rate of editorial publications and one-thirteenth the rate of community platforms. Analysis by DigitalApplied found that only 38% of AI citations come from pages ranking in the top 10 organic results — suggesting that AI engines pull from deeper content pools than traditional search. Sources with narrow content scope rarely produce enough citable material across enough query types to satisfy multiple engines' retrieval needs simultaneously.

How Machine Relations Measures This #

The Machine Relations Index reports citation rates: how often AI answer engines cite each source domain within each measured segment. The index measures six engines daily, publishes citation rates only after a segment clears the evidence floor (at least 10 observations across at least 7 distinct run dates), and assigns confidence tiers (A, B, C, or collecting) based on the volume of evidence behind each domain's rate.

Multi-engine citation overlap is a direct output of this measurement. Because the MRI tracks citations per engine, per domain, per segment, the number of engines that have cited any given domain is a direct observation — a simple count of which engines selected that source independently. This metric was not designed to measure overlap specifically — it falls out naturally from measuring citation rates across all six engines in parallel.

The public dataset reports citation rates, confidence tiers, and rankings. Source-level rankings show where a domain stands within its source type and across the full measured universe. For deeper context on how different source types earn citations at different rates, and how community platforms function as citation sources, see the linked MR research.

FAQ #

What percentage of sources get cited by multiple AI engines? #

It depends on how often the source is cited. Among domains cited at least ten times in 128 days, 98.5% are cited by two or more of the six engines and 15.4% by all six; among the 508 domains the Machine Relations Index grades, none is single-engine. Across the whole index, 67.4% of domains are single-engine, because 46.2% of the index was cited exactly once. Independent studies that count every cited domain once report a similar whole-population share.

Does appearing in more AI engines increase citation rates? #

The two move together, and the causation runs the other way. Domains cited by all six engines have a mean citation rate of 0.70% against 0.01% for single-engine domains. A domain's engine count cannot exceed its citation count, so being cited more is what makes multi-engine citation arithmetically available.

Which types of sources are most likely to be cited by multiple AI engines? #

Community and social platforms (33.3% reach 4+ engines), vendor-owned sources (25.5%), and editorial publications (20.3%) lead. The long tail of niche or unclassified domains reaches 4+ engines at only 6.7%. Source breadth, authority signals, and content volume across multiple topics all predict multi-engine citation.

Can you optimize content for multi-engine AI citation? #

Not through single-engine tactics. Each engine uses different retrieval, reranking, and citation selection logic. Empirical research on citation behavior confirms that source selection varies substantially across answer engines. The sources that reach all engines share structural characteristics — topical breadth, temporal consistency, established authority — rather than format tricks or keyword patterns. Building the kind of source that multiple engines independently select requires sustained citation-worthy content across enough subject areas to surface in diverse retrieval contexts.

Updated 2026-09-23: figures reflect the current Machine Relations Index methodology; see the MRI methodology and update log.