The Machine Relations Index publishes 96 category-and-question leaderboards in the release generated September 28, 2026. Reading off the domain ranked first in each one and checking how many of the six monitored engines cite that domain at all: 49 of the 96 winners are cited across all six. The other 47 hold the top spot in their segment while being absent from at least one engine outright, four of them missing three or more engines, and three of those four are missing Gemini, Google AI Mode and Google AI Overviews at the same time — the entire Google surface set — while still ranking first among every domain cited in their own segment.
What was measured #
Every figure here comes from the public release at machinerelations.ai/data/machine-relations-index.json, contract machine_relations_index_public_view_v2.0, methodology mri_score_v2.0, artifact 054b584266b8, generated September 28, 2026. The observation window runs May 10 to September 28, 2026 — 135 days, 24,639 cited source domains, 139,633 source events, six engines.
The six are ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity. Each publishes its own account of how it surfaces and links sources: OpenAI documents a search crawler distinct from its training crawler (OpenAI bots), Google describes AI Mode as its own search experience (Google, AI Mode) and documents its crawler fleet and AI-surface guidance separately (common crawlers, AI features and your site), Anthropic documents citations as part of Claude web search (Anthropic, web search tool), Google documents grounding with Search for Gemini (Gemini API), and Perplexity documents its own retrieval surface (Perplexity). Each reaches the web under its own user agent, governed by the Robots Exclusion Protocol as standardised in RFC 9309.
The methodology behind each leaderboard: a segment is one subject category paired with one buyer-question shape, and it publishes a ranked list only once it clears the evidence floor of at least 10 observed answer runs across at least 7 distinct run dates. This release carries 157 strata in the taxonomy, of which 151 are collectable and 96 are published. For each published segment, the domain with the highest citation rate holds rank 1. Separately, every domain in the Index carries its own overall engine-coverage record: the count and names of the six engines that were observed citing it at least once anywhere in the whole window, independent of category or question shape. Cross-referencing the two — which domain wins each segment, and how many engines cite that domain anywhere in the Index — produces the count in the lead.
Half of the winners are not universal #
| Engines that cite the segment winner | Number of segments (of 96) | Share |
|---|---|---|
| 6 of 6 | 49 | 51.0% |
| 5 of 6 | 11 | 11.5% |
| 4 of 6 | 32 | 33.3% |
| 3 of 6 | 3 | 3.1% |
| 2 of 6 | 1 | 1.0% |
Only the engine-count is being read here, not the identity of the winner: a domain wins a segment by out-citing every other domain that engines chose to cite for that category-and-question pair, and winning does not require appearing on every engine that made a choice. A segment can be won on the strength of four engines that agree heavily while two never observe the domain at all, and the leaderboard reports the winner without reporting the gap.
Which engine is missing most often #
Counting, across the 47 winners that miss at least one engine, which engine is the one doing the missing:
| Engine | Absent from this many of the 47 sub-universal winners |
|---|---|
| ChatGPT | 27 |
| Claude | 21 |
| Google AI Overviews | 18 |
| Google AI Mode | 14 |
| Gemini | 8 |
| Perplexity | 0 |
Perplexity cites every one of the 96 segment winners at least once somewhere in the Index; it is never the missing engine. ChatGPT is the most frequently absent, missing from 27 of the 47 non-universal winners, followed by Claude at 21. The pattern that recurs most often among four-engine winners is the same split documented separately for Reddit's own citation profile: cited by Gemini, Google AI Mode, Google AI Overviews and Perplexity, absent from ChatGPT and Claude. Twenty-one of the 32 four-engine winners in this release carry exactly that engine set, and reddit.com is the domain behind every one of those 21 — it wins the best-tools, how-buyers-choose, is-it-worth-it, problem-first-research and top-lists segments across several categories, always with the identical ChatGPT-and-Claude gap.
The narrowest winners are absent from the entire Google surface #
Four segment winners are cited by three or fewer of the six engines, the narrowest coverage in the release:
| Category and question | Winning domain | Engines that cite it | Missing |
|---|---|---|---|
| Enterprise Software, head-to-head comparisons | erpfocus.com | Claude, Perplexity | ChatGPT, Gemini, Google AI Mode, Google AI Overviews |
| Enterprise Software, top lists | procuredesk.com | ChatGPT, Claude, Perplexity | Gemini, Google AI Mode, Google AI Overviews |
| Fintech, is it worth it | paymentsandrisk.com | ChatGPT, Claude, Perplexity | Gemini, Google AI Mode, Google AI Overviews |
| Healthcare Services, head-to-head comparisons | adsc.com | ChatGPT, Claude, Perplexity | Gemini, Google AI Mode, Google AI Overviews |
Three of the four share the identical missing set: Gemini, Google AI Mode and Google AI Overviews, Google's entire AI-answer surface. Each of those three domains is, by this release's measure, the single most-cited source for its category-and-question pair — the answer an engine reaches for first when a buyer asks that exact question — while carrying no observed citation anywhere in the window from any Google AI surface. All four are specialist operator or trade sites rather than household names: ERPfocus, a procurement-software knowledge site; ProcureDesk, a procurement and AP-automation vendor; Payments & Risk, a payments-industry trade site; and Advanced Data Systems, a healthcare practice-management and billing vendor. A brand reading only "we rank first for our category" from a leaderboard position has no way to see that split; the segment table reports the winner, not which engines produced the win.
What this does not show #
This is a coverage count, not a caused-by-content claim: the Index measures which domains engines chose to cite, not why an engine's retrieval did or did not surface a given domain, and it draws no conclusion about content structure, page quality or a site's own retrievability. It also does not show whether the pattern is stable: a domain absent from Gemini in this window could appear in the next release once evidence accumulates further, and the evidence floor for a segment (10 runs, 7 dates) is a low bar for the segment to publish at all, not a claim that the winner's engine-coverage figure itself has settled. Engine-coverage is a whole-of-Index, whole-of-window figure for the domain, not a figure computed separately inside the one segment it happens to win; a domain could be cited by all six engines in other categories and still show this split in the one segment reported here if its citations there concentrate on fewer engines. Read the segment-level leaderboards and each domain's own record at machinerelations.ai/index before drawing a conclusion about one specific domain.