Which Publication Market I Trust

I trust the publications answer engines keep selecting, on many days, across engines, and I price everything else as a crawl or a distribution service until it shows the same recurrence. On the Machine Relations Index release I use for this decision, Medium is cited on 109 of 127 observed days and Forbes on 106, each across all six engines, at 5.04% and 4.15%. The wire and press-release distribution class, read as a class, is well under 1% of classified citation. Distribution volume does not buy the position at the top of that table.

The publication ranking behind those rates is Which Publications AI Search Engines Cite Most in 2026. That page is the deeper resource, the dataset. This essay is how I would use it as a founder: which column I would sort a budget by, why a one-engine leaderboard is the wrong market, and what would make me trust a wire release as a citation buy.

The measurement I would sort by

Release mri_score_v2.0+2026-09-20+47973f373a20, window May 10 to September 20, 2026, 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. Nothing receives a rate or a rank until it clears 10 observed runs across 7 distinct run dates.

Publication	Citation rate	Cited runs	Days cited	Engines	Confidence	Standing I would quote carefully
medium.com	5.04%	808	109 of 127	6	A	Rank 1 of 1,230 in the editorial class
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

I would quote the rate, the runs, the days, the engines, and the confidence before I quoted the class rank. The editorial-class rank is read off the Index's source-role field. In this release that field mixes source-type evidence with a placement catalog: 96 of 1,231 class members are admitted through the catalog, and editorial classification is moving toward source-type evidence alone. The class currently includes six platform domains beside edited newsrooms: medium.com at class rank 1, amazon.com at 7, prnewswire.com at 8, apple.com at 16, wordpress.com at 239, and blogspot.com at 330. Substack, Beehiiv, and Dev.to, the same kind of hosted 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%, including the top slot. The class denominator also moves between releases. Published figures carry 1,025 to 1,230 because each release counts the domains it classified. The rate, the runs, the days, the engines, and the confidence do not move when the class does. I would read the rate first, and I would not let "number one editorial publication" be the phrase that leaves the room without the host-versus-newsroom caveat.

PR Newswire is cited 1.11 times as often as TechCrunch, and both sit an order of magnitude below Medium and Forbes. Reuters, 13 cited runs on 13 days across 3 engines, does not clear the floor for a rank. Forbes is the second most-cited editorial publication on this denominator. I would buy recurrence at the top of the rate table. I would not buy a famous name that has 13 days out of 127 and call it the same market.

Two properties a leaderboard needs before I trust it

Engine breadth. A count from one index measures that index's crawl. A single retrieval index can be very good at finding syndicated releases. Six engines change the question from "what exists in a corpus" to "what gets chosen in an answer." I would not fund a target that one engine cites and five never do, unless the buyer I care about lives inside that one engine and I have said so.

The denominator. A real window rises and falls. A cumulative counter only climbs, and what it often counts is URLs entering a corpus. One press release becomes hundreds of syndicated URLs. One newsroom story becomes one URL. Count URLs in a crawl and syndication wins by construction. The Index counts answer runs over a fixed window. That is the market I trust, because it is the market a buyer actually sees.

Both properties explain the wire-class share. The 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 September 19, 2026 release. PR Newswire, the largest wire domain in the Index, is classified as an editorial publication in this release, so the nine-domain class excludes it. Counted with the wires, the class is ten domains holding 414 of those 112,518 cited runs, 0.37%. The finding holds either way: prnewswire.com's best standing in any published segment is number 15 of 448, so the class still holds none of the 880 top-ten slots. I would not brief 0.21% and 0.37% as a contradiction. I would brief them as the same fact on two denominators, and I would show the denominator.

Recurrence is selection, not syndication

A prestige placement is not automatically a retrieval asset. Recurrence here means being chosen again, on different questions, on different dates, by different engines. Medium on 109 of 127 days across six engines, Forbes on 106 across six, PR Newswire on 75, TechCrunch on 44, Reuters on 13. The top of the table is the source the engines return to. It is not automatically the biggest publisher.

"Be in more credible publications" is useful only when I can say which publications are actually selected. The per-segment tables publish that by question shape, with their own denominators and confidence grades. Density without a rate is how a company buys syndication and calls it visibility. I would rather have fewer outlets with a rate, a window, and an engine count than a long list of logos.

The engines document what they crawl and how to limit it: Google's common crawlers, OpenAI's bots, Perplexity's bots, and Anthropic's crawler policy. I would hold a citation claim to that standard of specificity. A vendor sentence I would keep next to the table, because positioning and standing are different facts: GlobeNewswire, read on September 20, 2026, sells a blueprint for AI-visible press releases, and globenewswire.com is cited in 74 of 16,039 monitored runs at confidence C.

The job I am describing, being the source a machine reaches for when it has to decide what is true, is the job I use Machine Relations to name. The measurement I would require before I believed an engine had resolved our name inside that source is entity resolution rate.

The decision

I would rank next quarter's publication targets by citation rate on this denominator, with the window, the 16,039-run count, engine breadth, days cited, and confidence on the same row. I would check engine breadth before I believed anyone else's leaderboard. I would read days-cited as the recurrence test: 100 of 127 days is selection, 13 of 127 is a burst. I would buy a wire release for the audiences and the disclosure obligations it already serves, and I would not buy it as an AI-visibility line. I would ask any vendor for the window, the denominator, and the evidence floor behind a citation number, and I would leave the number off the slide if those three are missing.

Scope: publication targeting for answer-engine citation, on release mri_score_v2.0+2026-09-20+47973f373a20 and the September 19 segment counts. It does not say Forbes is a bad human placement, and it does not say Medium's class rank is an earned-media instruction. The rate is what I trust. The class noun needs the caveat above. It does not transfer GlobeNewswire's marketing sentence into a forecast for a release we might send.

What would change my mind: a later release, on a stated window of similar length, in which the wire class took a material share of classified citation and held top-ten segment slots, or in which days-cited stopped separating the top of the table from the burst names. I would then price wires as citation. I would also trust a one-engine leaderboard if I had declared that engine to be the only buyer surface that mattered and the leaderboard disclosed its crawl. I would not trust a cumulative URL counter, because syndication wins that counter by construction.

FAQ

Which publications would I fund first for citation, on this release?

The ones engines keep selecting: Medium at 5.04%, 808 runs, 109 of 127 days, six engines, confidence A, and Forbes at 4.15%, 665 runs, 106 of 127 days, six engines, confidence A. I would fund them as retrieved domains with a rate, and I would keep Medium's editorial-class rank in the caveat column because the class includes hosts. TechCrunch at 1.02% and 44 days is a real but thinner selection. Reuters at 13 runs does not clear the rank floor.

Why do I not treat a wire release as the same purchase?

Because the nine-domain wire class is 232 of 112,518 classified cited runs, 0.21%, and 0 of 880 top-ten segment slots on the September 19 release. Including PR Newswire, which this release files outside that class, the ten domains are 414 runs, 0.37%, and PR Newswire's best segment standing is 15 of 448. GlobeNewswire's own domain is 74 of 16,039 runs at confidence C, beside a September 20, 2026 sales claim about AI-visible releases. I buy the wire for distribution. I buy citation where the days-cited column is heavy.

What would I reject in a vendor's citation chart?

A missing window, a missing denominator, or a missing evidence floor. A cumulative URL count. A single-engine crawl presented as the market. A class rank that hides the 96 catalog-admitted members and the six platform domains inside 1,358 of 13,525 editorial cited runs. The Index figures I am using were read from the public domain profiles on this release. I would ask a vendor to show me the same fields.

Does recurrence mean the outlet syndicates widely?

No. Recurrence means selection: different questions, different dates, different engines. Syndication multiplies URLs. The Index multiplies nothing. It counts an answer run in a fixed window of 127 days and 16,039 runs. That is why a newsroom story can outrank a wire that spawned hundreds of copies, and why I sort by the rate rather than by how many sites reposted the release.