Gemini generates more citations to elite information sources than any other AI engine. Across 17,540 source events tracked by the Machine Relations Index (MRI) spanning 6,020 domains, Gemini accounts for 30.3% of citations to top-ranked market and analyst sources — more than double ChatGPT's 6.7%. Independent tracking from Profound's 377,765-citation dataset confirms the pattern: Google Gemini produces 17.1% of total citations compared to ChatGPT's 4.9%. GeoXylia's 500-site benchmark found that AI platforms cite different sources for the same query in 73% of cases — the MRI data shows exactly how that divergence distributes across engines and source types.
Per-Engine Citation Volume for Elite Sources #
The MRI tracks citations across six AI engines — Perplexity, ChatGPT, Gemini, Claude, Google AI Mode, and Google AI Overviews — measuring how often each engine cites specific domains in response to buyer-intent queries across enterprise verticals.
Among six top-ranked sources (confidence grade A or B, each cited across at least 7 distinct measurement dates), the per-engine distribution reveals sharp differences:
| Source | Source Role | 30-Day Citations | Gemini | Perplexity | Google AI Mode | AI Overviews | Claude | ChatGPT |
|---|---|---|---|---|---|---|---|---|
| G2 | Market database | 145 | 52 (36%) | 30 (21%) | 29 (20%) | 16 (11%) | 9 (6%) | 9 (6%) |
| Crunchbase | Market database | 81 | 16 (20%) | 15 (19%) | 10 (12%) | 17 (21%) | 20 (25%) | 3 (4%) |
| Forbes | Analyst research | 65 | 29 (45%) | 5 (8%) | 10 (15%) | 10 (15%) | 2 (3%) | 9 (14%) |
| Deloitte | Analyst research | 50 | 11 (22%) | 16 (32%) | 5 (10%) | 7 (14%) | 4 (8%) | 7 (14%) |
| MarketsAndMarkets | Market database | 49 | 12 (24%) | 12 (24%) | 13 (27%) | 8 (16%) | 4 (8%) | 0 (0%) |
| Mordor Intelligence | Market database | 46 | 12 (26%) | 9 (20%) | 8 (17%) | 5 (11%) | 11 (24%) | 1 (2%) |
Aggregate across these sources: Gemini leads at 132 citations (30.3%), followed by Perplexity at 87 (20.0%), Google AI Mode at 75 (17.2%), Google AI Overviews at 63 (14.4%), Claude at 50 (11.5%), and ChatGPT at 29 (6.7%).
Data: Machine Relations Index, July 2026. 17,540 source events, 6,020 domains, MRI Score v1.1 (6-engine methodology).
Why Gemini Leads in Citation Volume #
Three structural factors explain Gemini's outsized citation share.
Search integration gives Gemini direct access to web sources. Gemini operates within Google's infrastructure and retrieves information from indexed web pages during response generation. As AI Citation Monitor noted, "Gemini grows faster and is wired into Google Search, so it cites the web more." Presenc AI's 2026 benchmark across 2,400 brands found that Gemini cites technical documentation at a 38% rate compared to ChatGPT's 30%, and original data/research at 33% compared to ChatGPT's 27% — confirming Gemini's preference for structured, factual content types that market databases provide.
Google's combined AI surfaces dominate the citation landscape. Across these six Elite sources, Google's three surfaces — Gemini, AI Mode, and AI Overviews — account for 270 of 436 citations (62%). Independent data from Profound's tracking of 377,765 citations across eight engines shows a consistent pattern: Google AI Mode leads at 23.3%, Gemini follows at 17.1%, and Google AI Overviews contributes 14.6%. Combined, Google produces over 55% of all tracked citations.
ChatGPT's retrieval architecture limits its citation output. Foglift Research found that ChatGPT cites the vendor's own first-party site 68% of the time across 2,583 citations in their Q2 2026 benchmark. This preference for first-party content means ChatGPT draws from a narrower pool of external sources. The MRI data confirms this: ChatGPT produced zero citations to MarketsAndMarkets and just one to Mordor Intelligence across the full measurement window.
Engine-Specific Source Preferences #
The per-source breakdowns reveal that each engine has distinct citation preferences — a finding that LLM Pulse confirmed across their 5.3 million citation dataset spanning 470,380 AI answers.
Claude concentrates citations on Crunchbase. Of Claude's 50 total citations to these six sources, 20 (40%) go to Crunchbase alone. No other engine shows this level of source concentration. Claude gives Crunchbase the highest citation share of any engine-source pair in the dataset.
Perplexity distributes citations more evenly but retrieves more sources per answer. Perplexity's citation share ranges from 8% (Forbes) to 32% (Deloitte) across these sources, with no single source receiving more than a third of its citations. Attrifast's vertical citation study found that Perplexity cites a median of 6.4 unique domains per answer — 2.1x more than ChatGPT (3.1) and 2.7x more than Gemini (2.4). Perplexity's higher source breadth per answer is consistent with its six-stage citation pipeline, even though its total citation volume trails Gemini's because it handles fewer queries.
Google AI Mode favors market data providers. Google AI Mode gives MarketsAndMarkets its highest share (27%) of any source in the dataset and maintains consistent citation rates across market databases. This aligns with AI Mode's function as a research assistant within Google Search, where structured market sizing data serves buyer-intent queries directly.
As AuthorityTech's publication-level analysis found, a placement in the wrong outlet relative to a specific engine can carry 4.5x less answer influence than one in the right outlet. The MRI data now quantifies that gap at the source level.
What This Means for Source Visibility Strategy #
The engine-level divergence creates a specific strategic problem: optimizing for "AI search" as a single channel ignores the fact that each engine selects from a different citation pool. QuickSEO's cross-engine analysis documented the same pattern: ChatGPT, Claude, Gemini, and Perplexity each cite different publications for identical queries.
Source owners need engine-specific visibility data. A domain with a strong overall citation rate may still be invisible to specific engines. MarketsAndMarkets carries 49 total citations and a confidence B grade — but ChatGPT has never cited it. Brands appearing on MarketsAndMarkets reports get zero ChatGPT referral traffic from that placement regardless of the domain's overall MRI performance.
Google's AI surfaces are the volume play. For sources seeking maximum citation reach, Google's three surfaces produce 62% of citations in this dataset. The Profound tracking data confirms this at scale: Google AI Mode alone accounts for 87,921 of 377,765 total citations (23.3%).
ChatGPT requires a different approach. ChatGPT's low citation volume to external sources means brands relying on third-party mentions in market databases will see less ChatGPT visibility. The citation absorption gap — where a source appears in the retrieval list but contributes nothing to the response text — is most pronounced in ChatGPT's architecture. The Boring SEO's comparison documented how ChatGPT's response construction differs structurally from Perplexity and Gemini, consistent with the MRI's finding that ChatGPT accounts for under 7% of citations to elite sources.
Vertical concentration amplifies the engine gap. Attrifast's citation-by-vertical study found that SaaS queries generate the highest citation density (5.1 unique domains per answer), while healthcare concentrates on a small set of high-trust sources where the top 10 domains capture 71.2% of citation slots. This means the engine-level divergence documented in the MRI data is even more pronounced in verticals with high source concentration — a domain's engine coverage matters most where the citation pool is narrow.
How the Machine Relations Index Measures Engine-Level Citations #
The Machine Relations Index v2 measures source-segment citation rates — how often AI answer engines cite each source domain — across six engines: Perplexity, ChatGPT, Gemini, Claude, Google AI Mode, and Google AI Overviews. The July 2026 dataset covers 17,540 source events across 6,020 unique domains. A segment publishes a citation rate only after clearing an evidence floor of at least 10 observations across at least 7 distinct measurement dates. Domains are graded into confidence tiers (A, B, C, or collecting) based on the volume of evidence behind each rate.
The six sources profiled in this analysis all carry confidence grades A or B, with citation counts ranging from 145 (G2, #1 among 307 market database sources) to 46 (Mordor Intelligence). All six are cited by every engine except MarketsAndMarkets, which ChatGPT does not cite in the current measurement window.
Engine-level citation data is derived from the MRI's per-event attribution, which records which engine produced each citation, the query that triggered it, the vertical classification, and the source's position within the response. The per-engine breakdowns in this analysis reflect 30-day rolling citation counts. The Searchless Journal's 500-query benchmark uses a similar multi-engine measurement framework to evaluate citation behavior across AI search systems.
FAQ #
Which AI engine produces the most citations? #
Gemini produces the most citations to top-ranked market and analyst sources. In the MRI dataset of 17,540 source events, Gemini accounts for 30.3% of citations to the six highest-ranked sources. Independent tracking from Profound confirms Google Gemini generates 17.1% of all tracked citations, second only to Google AI Mode at 23.3%.
Why does ChatGPT cite fewer external sources than other engines? #
ChatGPT's retrieval architecture favors first-party vendor content. Foglift Research's Q2 2026 benchmark found ChatGPT cites the vendor's own site 68% of the time. This reduces the share of citations going to third-party sources like market databases and analyst firms.
Does Gemini's citation volume advantage affect brand visibility? #
Yes. Brands that appear on sources frequently cited by Gemini — particularly G2 and Forbes — receive disproportionate citation exposure. G2 receives 36% of its 145 total citations from Gemini alone. Brands absent from Gemini-favored sources lose access to the largest single-engine citation channel in the current AI search environment.
How should source owners respond to engine-level citation differences? #
Track citation performance per engine rather than in aggregate. A source with a strong overall citation rate may still have zero citations from a specific engine. The MRI's per-engine breakdown identifies these gaps so source owners can prioritize visibility on surfaces that reach their target engines.
Last updated: July 25, 2026