Forbes Outpaces PR Newswire 3.65x in AI Citations: What the Machine Relations Index Shows

The AI citation market has moved away from wires and toward editorial. On one denominator, forbes.com is cited in 665 of 16,039 monitored answer runs and prnewswire.com in 182 of the same 16,039, so Forbes is cited 3.65 times as often. The whole wire and press-release distribution class takes well under 1% of classified citation.

What the Index measures

Every figure below comes from the Machine Relations Index release mri_score_v2.0+2026-09-20+47973f373a20, generated today. The window is 2026-05-10 to 2026-09-20, 127 days, 919 monitored prompts, 16,039 answer runs, 22,320 cited source domains, six engines: ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity. A domain is counted once per answer run in which it is cited. Rates publish only after a segment clears the evidence floor of 10 observed runs across 7 distinct run dates, and every domain carries a confidence grade for the weight of evidence behind it.

The leading publications, on that one denominator:

Publication	Citation rate	Cited runs	Days cited	Engines	Confidence	Standing
medium.com	5.04%	808 of 16,039	109 of 127	6	A	rank 1 of 1,230 classified editorial publications
forbes.com	4.15%	665 of 16,039	106 of 127	6	A	rank 2 of 1,230
techradar.com	2.87%	460 of 16,039	—	4	A	rank 3 of 1,230
yahoo.com	1.93%	310 of 16,039	—	5	A	rank 4 of 1,230
prnewswire.com	1.13%	182 of 16,039	75 of 127	6	B	rank 8 of 1,230
techcrunch.com	1.02%	164 of 16,039	44 of 127	5	B	rank 9 of 1,230

Medium is first among editorial publications, at a 5.04% citation rate with Confidence A and presence on 109 of 127 observed days. If you are choosing where to put a piece of writing on the assumption that engines reach for it, Medium is the strongest single answer in this class.

The wire layer sits roughly an order of magnitude below the top of the editorial class.

The wire layer, measured

The Index classifies every cited domain into a source role. Wire and press-release distribution is one of them, and it is the smallest class that gets cited at all: 9 domains holding 232 cited runs out of 112,518 across the nine classes, which is 0.21% of classified citation.

Wire domain	Citation rate	Cited runs	Days cited	Engines	Confidence	Standing
businesswire.com	0.61%	98 of 16,039	51 of 127	6	C	rank 74 of 22,320 observed domains
globenewswire.com	0.46%	74 of 16,039	37 of 127	5	C	rank 110 of 22,320

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.

The deeper reading is in AuthorityTech's analysis of where top-ten citation slots actually go. Across all 88 published segments on the 2026-09-19 release, wire distribution held 0 of 880 top-ten slots — the full detail is in how AI search engines decide what to cite.

Why rate, not count, is the measure

A cumulative counter rewards whoever published the most URLs earliest, which is exactly what a wire does: one release, hundreds of syndicated copies, each one a separate URL entering the count. Ranked that way, wires win by construction and the result says nothing about whether an engine chose them. Ranked as a rate, the share of monitored answer runs in which a domain is cited at all, the outlets that get selected rather than syndicated move to the top.

That is the difference between counting inventory and measuring selection. A trailing window should move up and down; a number that only climbs under a window label is a counter accumulating since it was switched on.

The Index publishes its window, its denominator, its evidence floor and its release id on every page, so every figure on it is checkable by anyone.

What the wire industry says about itself

None of the above is a claim that wires are selling something they know to be worthless. They are selling AI visibility, explicitly and in their own words, and today those words are easy to read.

GlobeNewswire's home page, read 2026-09-20, invites you to "shape your presence in AI answers" and leads with the SOAR Content Framework, described there as "The blueprint for AI-visible press releases," built on "one of the largest AI visibility studies in the industry," alongside an AI Press Release Optimizer that "improves visibility in answer-engine responses." PR Newswire's home page, read the same day, carries a Create with AI product and announces that Cision has acquired Trajaan, described there as a search intelligence and GenAI monitoring platform.

Set that beside the measurement. globenewswire.com is cited in 74 of 16,039 monitored answer runs, on 37 of 127 days, at Confidence C. That is the whole comparison, and it is a comparison of stated positioning against measured standing, not a claim about anyone's intent.

The decision, recorded

Wire distribution is a distribution cost, not an AI visibility line. I am not buying it as one, and I would not advise a founder to.

It is the same shape as the decision I recorded yesterday about international press coverage: a spend the market sells as an AI visibility input, which measures as something else entirely. The pattern is not that the vendors are lying. It is that the whole category was ranking on volume, because volume was the only number anyone had.

The reversal conditions are specific, and the next release can fire them:

1. Any wire domain reaching Confidence B or better with a citation rate above 2% on the whole-index denominator.
2. The wire and press-release distribution class clearing 1% of classified citation, counted with prnewswire.com in it — 0.37% today.
3. Any wire domain taking a top-ten slot in a published segment, against the 0 of 880 it holds today.

If any of those hold on a future release, this entry will record it on this page.

What to do instead

1. Rank publication targets by measured citation rate on one denominator, never by citation counts from different windows. Every publication in the Index carries its rate, its run count, its days cited and its grade on one page.
2. Ask any vendor quoting you a citation number for its window, its denominator and its evidence floor. If the answer is a raw count with no denominator, you are looking at inventory.
3. Control what the engines can actually read from you. The crawlers are documented by the engines themselves: Google's common crawlers and its AI features guidance, OpenAI's bots, Perplexity's bots, and Anthropic's crawler policy.
4. Treat the shift as durable rather than seasonal. Pew's research on how Americans view AI tracks a public that is cautious but increasingly using these systems day to day, which is the demand side of the same curve.

The thesis of this post is that the citation market is rotating from wires toward editorial. On the Index's measurement, it has already rotated. Machine Relations is the frame I use for that: media as retrieval signal and citation leverage rather than exposure. Operators who treat distribution as an input to authority rather than an output, like Christian Lehman, were reading it correctly early.

Sources and method

Index figures read live 2026-09-20 from machinerelations.ai/index and the individual domain profiles linked in the tables above, on release mri_score_v2.0+2026-09-20+47973f373a20, window 2026-05-10 to 2026-09-20. Wire vendor positioning was read live 2026-09-20 from each vendor's own site and is quoted, not paraphrased. Segment top-ten counts are from the 2026-09-19 release as published in AuthorityTech's analysis. If you want your own domain's standing on the same denominator, the visibility audit runs it, and the neutral profile is free to read at machinerelations.ai.

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