Machine Relations Index

Machine Relations Index (MRI)

Name: Machine Relations Index
Published: 2026-05-10
License: https://creativecommons.org/licenses/by/4.0/

A neutral public benchmark for how AI answer engines cite sources, built from observed root-domain citations across six engines. MRI reports citation rates by category and question type once minimum sample thresholds are met.

28,631 domains175,782 source events22,381 answer runs112/151 measurable segments validated

Measurement window2026-05-10 to 2026-10-09

Published 2026-10-09 from the current dataset release covering 28,631 cited domains and 175,782 source events.

Sample threshold10/7

Rates publish after enough answer runs across distinct observation dates.

Segment coverage74.2%

112 of 151 measurable segments validated, 39 still collecting observations. A further 6 of the 157 segments in the full universe cannot accumulate evidence and are excluded from every rate.

Citation leaders

Top cited domains

Source universe

Cited source index

Initial view shows the top 50 domains from 28,631 total.

Data through 2026-10-09

Row	Source	Class			Confidence
1	Redditreddit.com · #1 of 28,631 observed domains · source site	Community and social platform	10.52%	2,355	Confidence A
2	YouTubeyoutube.com · #2 of 28,631 observed domains · source site	Search or media platform	7.54%	1,687	Confidence A
3	LinkedInlinkedin.com · #3 of 28,631 observed domains · source site	Community and social platform	6.06%	1,356	Confidence A
4	Forbesforbes.com · #1 of 1669 classified editorial publications · source site	Editorial publication	4.25%	951	Confidence A
5	Mediummedium.com · #5 of 28,631 observed domains · source site	Community and social platform	4.07%	910	Confidence A
6	TechRadartechradar.com · #2 of 1669 classified editorial publications · source site	Editorial publication	2.89%	647	Confidence A
7	Arxiv.orgarxiv.org · #7 of 28,631 observed domains · source site	Academic and government source	2.82%	631	Confidence A
8	G2g2.com · #8 of 28,631 observed domains · source site	Market and company database	2.69%	602	Confidence A
9	NIHnih.gov · #9 of 28,631 observed domains · source site	Academic and government source	2.69%	602	Confidence A
10	Gartner.comgartner.com · #10 of 28,631 observed domains · source site	Analyst and consulting research	2.33%	521	Confidence A
11	Yahoo Newsyahoo.com · #3 of 1669 classified editorial publications · source site	Editorial publication	2.16%	483	Confidence A
12	Healthlinehealthline.com · #4 of 1669 classified editorial publications · source site	Editorial publication	1.85%	414	Confidence B
13	Microsoft.commicrosoft.com · #13 of 28,631 observed domains · source site	Vendor-owned source	1.84%	411	Confidence A
14	Amazon.comamazon.com · #14 of 28,631 observed domains · source site	Other observed source	1.66%	372	Confidence A
15	NerdWalletnerdwallet.com · #5 of 1669 classified editorial publications · source site	Editorial publication	1.61%	361	Confidence A
16	Ibm.comibm.com · #16 of 28,631 observed domains · source site	Vendor-owned source	1.58%	354	Confidence A
17	TechTargettechtarget.com · #6 of 1669 classified editorial publications · source site	Editorial publication	1.48%	332	Confidence A
18	Guideflow.comguideflow.com · #18 of 28,631 observed domains · source site	Vendor-owned source	1.35%	302	Confidence A
19	Landbase.comlandbase.com · #19 of 28,631 observed domains · source site	Vendor-owned source	1.34%	301	Confidence A
20	Paloaltonetworks.compaloaltonetworks.com · #20 of 28,631 observed domains · source site	Vendor-owned source	1.28%	286	Confidence B
21	Substack.comsubstack.com · #21 of 28,631 observed domains · source site	Community and social platform	1.27%	285	Confidence B
22	Dev.todev.to · #22 of 28,631 observed domains · source site	Community and social platform	1.25%	279	Confidence B
23	Crunchbase.comcrunchbase.com · #23 of 28,631 observed domains · source site	Market and company database	1.24%	278	Confidence B
24	Wikipediawikipedia.org · #24 of 28,631 observed domains · source site	Academic and government source	1.13%	252	Confidence B
25	Facebook.comfacebook.com · #25 of 28,631 observed domains · source site	Community and social platform	1.10%	246	Confidence B
26	Sentinelone.comsentinelone.com · #26 of 28,631 observed domains · source site	Vendor-owned source	1.09%	245	Confidence B
27	The New York Timesnytimes.com · #7 of 1669 classified editorial publications · source site	Editorial publication	1.09%	244	Confidence B
28	Github.comgithub.com · #28 of 28,631 observed domains · source site	Community and social platform	1.02%	228	Confidence B
29	Hubspot.comhubspot.com · #29 of 28,631 observed domains · source site	Vendor-owned source	1.01%	225	Confidence B
30	CNBCcnbc.com · #8 of 1669 classified editorial publications · source site	Editorial publication	0.97%	217	Confidence B
31	Alibaba.comalibaba.com · #31 of 28,631 observed domains · source site	Other observed source	0.93%	209	Confidence B
32	PR Newswireprnewswire.com · #32 of 28,631 observed domains · source site	Wire and press-release distribution	0.93%	208	Confidence B
33	PCMagpcmag.com · #9 of 1669 classified editorial publications · source site	Editorial publication	0.92%	206	Confidence B
34	People Managing Peoplepeoplemanagingpeople.com · #10 of 1669 classified editorial publications · source site	Editorial publication	0.92%	206	Confidence B
35	Experian.comexperian.com · #35 of 28,631 observed domains · source site	Vendor-owned source	0.91%	204	Confidence B
36	Deloitte.comdeloitte.com · #36 of 28,631 observed domains · source site	Analyst and consulting research	0.91%	203	Confidence B
37	Zapier.comzapier.com · #37 of 28,631 observed domains · source site	Vendor-owned source	0.90%	202	Confidence B
38	Stripe.comstripe.com · #38 of 28,631 observed domains · source site	Vendor-owned source	0.88%	196	Confidence B
39	Rippling.comrippling.com · #39 of 28,631 observed domains · source site	Vendor-owned source	0.86%	192	Confidence B
40	Erpresearch.comerpresearch.com · #40 of 28,631 observed domains · source site	Market and company database	0.85%	190	Confidence B
41	Bankrate.combankrate.com · #41 of 28,631 observed domains · source site	Other observed source	0.84%	189	Confidence B
42	Grandviewresearch.comgrandviewresearch.com · #42 of 28,631 observed domains · source site	Market and company database	0.84%	189	Confidence B
43	Harvard.eduharvard.edu · #43 of 28,631 observed domains · source site	Academic and government source	0.84%	188	Confidence B
44	Nvidia.comnvidia.com · #44 of 28,631 observed domains · source site	Vendor-owned source	0.79%	176	Confidence B
45	Databricks.comdatabricks.com · #45 of 28,631 observed domains · source site	Vendor-owned source	0.77%	172	Confidence B
46	TechCrunchtechcrunch.com · #11 of 1669 classified editorial publications · source site	Editorial publication	0.76%	170	Confidence B
47	Y Combinatorycombinator.com · #47 of 28,631 observed domains · source site	Community and social platform	0.76%	170	Confidence B
48	Apple Newsapple.com · #48 of 28,631 observed domains · source site	Other observed source	0.74%	165	Confidence B
49	WebMDwebmd.com · #12 of 1669 classified editorial publications · source site	Editorial publication	0.73%	163	Confidence B
50	Worldmetrics.orgworldmetrics.org · #50 of 28,631 observed domains · source site	Analyst and consulting research	0.73%	163	Confidence B

Agent access

Use this data in your agent

Every figure on this index can be read by machines. The MCP endpoint and the REST API return the same domain, category and segment records you see here. Each record carries its release identifier, window and denominators, and segment records also carry the evidence floor, so an agent can cite the exact number and when it was measured. The data is published under CC BY 4.0.

Read a segment rank as a statement about the questions behind it, not about a whole market. Some segments rest on only a few questions, and published rates still include runs where the question names the domain's own brand. Pulling from the API reproduces the published numbers; it does not re-run the measurement.

Read the returned questionContext, questionSet and naming fields before interpreting a rank. Missing or null context means unavailable for that release, never zero. When present, read questionContext.readAs and limitations, including namesReviewedCoverage and sourceRoleBasis; a rank describes the sampled questions, not an entire market.

MCP endpoint: https://machinerelations.ai/mcp

The Machine Relations Index measures how often AI answer engines cite each source domain when people ask real buying and research questions. It observes six engines — Perplexity, ChatGPT, Gemini, Claude, Google AI Mode, and Google AI Overviews — across a fixed basket of 24 measured subject categories plus a legacy news-topic bucket (25 taxonomy nodes in all). Findings are grouped into strata, where a stratum is one subject category paired with one question type. A stratum publishes a citation rate only after it clears the evidence floor of at least 10 observed runs across at least 7 distinct dates; below that line it is marked collecting rather than scored, so early or thin signal is never shown as settled. Days on which an engine was asked at least five times and returned no citation at all are treated as days that engine was not collecting, and their runs are excluded from the denominator: a run that cannot reach the numerator would otherwise depress the rate it is counted in. Every excluded engine-day is listed, with its run count, in the denominator_exclusion block of each release. Every domain an engine cites is measured on equal terms — membership in any external catalog is recorded only as a label and is never used to decide what gets scored. The public Index reports rates and rankings; it excludes raw cited URLs, internal query identifiers, and provider payloads. Validated source segments currently cover 74.2% of the measurable benchmark, which is the 151 segments that can accumulate evidence out of 157 in the full stratum universe.

Release mri_score_v2.1+2026-10-09+cae1a3ac9025. Degraded coverage: 5 of 6 answer engines are currently returning answers. gemini quiet (14 of 115 collected calls returned an answer, 12%). Coverage across the window is uneven: gemini answered on 124 of 146 observed days (2,011 of 5,116 returned runs produced a citation); google_ai_mode answered on 104 of 146 observed days (1,371 of 1,735 returned runs produced a citation); google_ai_overviews answered on 87 of 146 observed days (774 of 1,218 returned runs produced a citation). Engines below 90% of the window are named here; per-engine day and run counts for all 6 are in engine_roster.

Machine Relations Index (MRI)

Top cited domains

Cited source index

Use this data in your agent

Check how AI systems cite your brand.