The Publication Market AI Actually Trusts: What the Machine Relations Index Shows Across Six Engines

Which publications does AI actually trust? Measured across six engines, the answer is the sources engines keep coming back to: Medium is cited on 109 of 127 observed days and Forbes on 106, each across all six engines, at citation rates of 5.04% and 4.15%. Distribution volume does not buy that position. The wire and press-release distribution class takes well under 1% of classified citation.

What the measurement says

Machine Relations Index release mri_score_v2.0+2026-09-20+47973f373a20, window 2026-05-10 to 2026-09-20, 127 days, 919 monitored prompts, 16,039 answer runs, 22,320 cited domains, six engines. A domain counts once per answer run in which it is cited, engine breadth is published per domain, and nothing gets a rate or a rank until it clears an evidence floor of 10 observed runs across 7 distinct run dates.

Publication	Citation rate	Cited runs	Days cited	Engines	Confidence	Measured standing
medium.com	5.04%	808	109 of 127	6	A	rank 1 of 1,230 editorial publications
forbes.com	4.15%	665	106 of 127	6	A	rank 2 of 1,230
prnewswire.com	1.13%	182	75 of 127	6	B	rank 8 of 1,230
techcrunch.com	1.02%	164	44 of 127	5	B	rank 9 of 1,230
entrepreneur.com	0.11%	18	13 of 127	4	Collecting	below the floor for a rank
reuters.com	0.08%	13	13 of 127	3	Collecting	below the floor for a rank

How to read the editorial-class rank. The editorial-class rank above is read off the Machine Relations Index's source_role field, which in the current release combines source-type evidence with AuthorityTech's placement catalog; 96 of the 1,231 class members are admitted through the catalog, and the Index is moving editorial classification to source-type evidence alone. The class currently includes six platform domains alongside edited newsrooms: medium.com (class rank 1), amazon.com (7), prnewswire.com (8), apple.com (16), wordpress.com (239) and blogspot.com (330). At the same time substack.com, beehiiv.com and dev.to — the same hosted open-publishing product as Medium — are filed as community platforms. Those six hold 1,358 of the editorial class's 13,525 cited runs, 10.04%, and they include the class's top slot. The class denominator also moves between releases (published figures carry 1,025 to 1,230) because it counts the domains classified in each release. What is unaffected: the citation rate, cited runs, days cited, engine breadth and confidence on this page are direct per-domain observations over a fixed run set, and are the primary measure; read the rate first.

PR Newswire is cited 1.11 times as often as TechCrunch, and both sit an order of magnitude below Medium and Forbes. Reuters, at 13 cited runs on 13 days across 3 engines, does not yet clear the floor for a rank. Forbes is the second most-cited editorial publication on this denominator.

Two things a trustworthy leaderboard has to get right

The first is engine breadth. A count assembled from one index measures that index's crawl, and a single retrieval index can be very good at finding syndicated press releases. Ask six engines and the question stops being "what exists in a corpus" and becomes "what gets chosen in an answer."

The second is the denominator. A real window rises and falls. A cumulative counter only climbs, and what it counts is URLs entering the corpus. The Index counts answer runs over a fixed window.

Both matter for the same reason. One press release becomes hundreds of syndicated URLs; one newsroom story becomes one. Count URLs in a crawl and syndication wins by construction. Count answer runs across six engines and the whole wire and press-release distribution class (nine domains) takes 232 cited runs of 112,518 across the nine source classes, 0.21% of classified citation, with 0 of 880 top-ten segment slots on the 2026-09-19 release. One note so two numbers on this page do not read as contradicting each other: prnewswire.com, the largest wire domain in the Index, is classified as an Editorial publication in this release rather than in the wire class, so the nine-domain class above excludes it. Counted with the wires, the class is ten domains holding 414 of those 112,518 cited runs, or 0.37% of classified citation. The finding does not move: prnewswire.com's best standing in any published segment is #15 of 448, so the class still holds 0 of 880 top-ten slots, and its rank 8 and the class share simply sit on two different denominators.

Humans confer status. Machines confer retrievability.

A prestige placement is not automatically a retrieval asset, earned media works as infrastructure rather than as a trophy case, and recurrence beats a single shiny moment.

Recurrence here means breadth of selection, not breadth of distribution: being chosen repeatedly, on different questions, on different dates, by different engines. Medium is cited on 109 of 127 observed days across all six engines and Forbes on 106 across all six. PR Newswire manages 75 days, TechCrunch 44, Reuters 13. The top of that table is not the biggest publisher. It is the source engines keep coming back to.

That is why "appear across a denser network of credible, indexable publications" is only useful if you can tell which ones are actually selected, and the per-segment tables publish that per question shape, with their own denominators and confidence grades. Aiming at density without a rate is how you end up buying syndication and calling it visibility.

What to do

1. Rank targets by citation rate on one denominator, with window, run count, engine breadth and confidence attached. The Index publishes all of it, free to read.
2. Check engine breadth before you believe a leaderboard. A source cited by one engine is a crawl artifact, not a market.
3. Read days-cited. A source on 100 of 127 days is being selected; a source on 13 is catching bursts.
4. Treat a wire release as distribution to its original audiences and price it there.
5. Ask any vendor for the window, the denominator and the evidence floor behind any citation number.

The engines document what they crawl and how to control it: Google's common crawlers, OpenAI's bots, Perplexity's bots, Anthropic's crawler policy. That is the standard of specificity a citation claim should meet. Meanwhile GlobeNewswire, read 2026-09-20, sells "the blueprint for AI-visible press releases"; globenewswire.com is cited in 74 of 16,039 monitored runs at Confidence C. Positioning and standing are separate facts.

That is where entity resolution rate and Machine Relations stop being theory. The job is not to get covered, and it is not to be crawlable. It is to be the source a machine reaches for when it has to decide what is true.

Sources and method

Index figures read live 2026-09-20 from machinerelations.ai/index and the linked domain profiles on release mri_score_v2.0+2026-09-20+47973f373a20. Segment top-ten counts come from the 2026-09-19 release as published in AuthorityTech's analysis. Vendor positioning was read live on 2026-09-20 and is quoted, not paraphrased. Your own domain's standing on the same denominator is available through the visibility audit.

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