The Editorial Leaderboard AI Actually Uses: Medium and Forbes Lead in the Machine Relations Index

Strip out the distribution surfaces and look at what engines actually reuse, and the editorial leaderboard is not the one most founders assume. Measured on one denominator, Medium ranks first, Forbes is cited four times as often as TechCrunch, and Reuters does not yet clear the evidence floor.

The leaderboard

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, and nothing publishes 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
techcrunch.com	1.02%	164	44 of 127	5	B	rank 9 of 1,230
businessinsider.com	0.62%	100	41 of 127	5	B	rank 20 of 1,230
fortune.com	0.37%	59	28 of 127	4	C	rank 36 of 1,230
reuters.com	0.08%	13	13 of 127	3	Collecting	below the floor for a rank
prnewswire.com	1.13%	182	75 of 127	6	B	rank 8 of 1,230

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.

TechCrunch is ninth, and Forbes is cited in 665 runs against its 164. Medium, often filed under "open publishing, weak editorial filtering" and set aside as a volume surface, is the single most-cited editorial publication on this denominator, on 109 of 127 observed days across all six engines.

Why a rate, not a count

A leaderboard is only as good as its denominator. A cumulative counter that only climbs is counting from the beginning, and what it counts is URLs published, not answers won. That does not only inflate the wire. It scrambles an editorial table too, because a newsroom that publishes more URLs outranks one that publishes fewer regardless of what engines select. The Index counts answer runs over a fixed window, which is why Forbes sits above TechCrunch here.

This is the same mistake founders make in their own lives.

They confuse abundance with authority.

A hundred meetings feels like momentum. It can just be avoidance. A packed calendar feels important. It can just be a shield against the one hard decision you do not want to make.

PR has the same trap. More placements. More links. More noise. Then the buyer asks ChatGPT, Gemini, Claude, or Perplexity a real buying question and the answer resolves through a publication you never earned because you optimized for count instead of consequence.

That is why the distinction matters.

Syndication still has a job. It is just not this one. 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, and held 0 of 880 top-ten segment slots on the 2026-09-19 release. One note on reading that figure: the Index files prnewswire.com, the largest wire domain in it, as an editorial publication, so the nine-domain class leaves it out. Counted with the wires, the class is ten domains at 414 cited runs of 112,518, or 0.37% of classified citation. It holds regardless: prnewswire.com's best standing in any published segment is #15 of 448, so the class still holds 0 of 880 top-ten slots. Press release surfaces are structured and crawlable; being crawlable is not being chosen. Conductor's 2026 AEO/GEO benchmarks report made the macro shift explicit: AI is replacing the website as the first place many buyers encounter a brand.¹ Muck Rack's Generative Pulse reporting pushed the same direction from another angle: over four-fifths of tracked AI citations came from earned media sources, while journalism becomes even more dominant when the query implies freshness.²

But syndication and editorial authority are not interchangeable.

One expands the surface.

The other changes the answer.

That split is the whole point of Machine Relations as a marketing discipline. The machine does not reward you for existing. It rewards the sources it trusts to compress reality on its behalf. If you want the category map, the top publications cited by AI search engines in B2B now makes that hierarchy visible. If you want the broader system, the Machine Relations Stack explains why distribution, entity clarity, citation architecture, and measurement are different jobs.

And this is where most operators break.

They want one metric.

One clean number.

One dashboard that tells them they are winning.

Life is rarely that generous.

The external problem is usually an internal problem wearing a mask. If you are chasing raw mention volume, there is usually a deeper need underneath it. Validation. Simplicity. The relief of not having to make a harder judgment call. Raw counts let you pretend all citations are equal because that is emotionally easier than admitting some surfaces matter far more than others.

They don't.

The concentration is real, and it is visible in a column a raw counter cannot produce: days cited. Medium is cited on 109 of 127 observed days and Forbes on 106, both across all six engines. TechCrunch manages 44 days across five, Business Insider 41, Fortune 28, Reuters 13. The machine is tightening around a narrow set of publishers it returns to repeatedly, across different questions on different dates. That is recurrence, and it is the thing volume was standing in for.³

What matters in 2026 is not whether a placement generated a temporary traffic spike.

What matters is whether it becomes reusable memory inside the answer layer.

A mention that sends a few hundred humans to your site but never gets reused by AI is a different asset from a placement on a publication the models cite over and over for buyer-intent questions.

One gives you attention.

The other becomes memory.

That is the distinction I care about.

Because the founder's real job is not to collect activity.

It is to shape the environment that makes future decisions easier and more favorable.

You do that in your company the same way you do it in your own head.

You stop treating every signal as equal.

You stop rewarding noise because it is measurable.

You start asking which inputs the system actually trusts when the moment of judgment arrives.

That is the work.

Everything else is motion theater.

If you want the tactical version of this for operators, AuthorityTech's breakdown of which publications get cited most by AI search engines in 2026 is the cleaner field guide. If you want the first-person context around how I think about these shifts, start with Why I Coined Machine Relations. If you want the human layer behind the same thesis, Christian Lehman's writing tracks how narrative leverage compounds once the right third-party surfaces start carrying the story. And if you want to see whether your own brand is showing up inside the answer layer yet, run an AI visibility audit.

If it is not, the problem is probably not content volume.

It is that you are still feeding the machine the wrong kind of proof.

Sources and method

The engines publish what they crawl and how to control it: Google's common crawlers, OpenAI's bots, Perplexity's bots, Anthropic's crawler policy. A citation claim should meet that same standard of specificity: a rate with a denominator, a window, an engine count and a confidence grade attached. The third-party sources on this page, Conductor and Muck Rack, are linked at their own sources.

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

Notes

Footnotes

1. Conductor, "The 2026 AEO / GEO Benchmarks Report," accessed March 30, 2026, https://www.conductor.com/academy/aeo-geo-benchmarks-report/. ↩

2. Muck Rack reporting cited via AuthorityTech synthesis and Generative Pulse coverage summarized March 2026; see also AuthorityTech, "Which Publications Get Cited Most by AI Search Engines in 2026," https://authoritytech.io/blog/which-publications-get-cited-most-ai-search-engines-2026. ↩

3. Machine Relations Index, release mri_score_v2.0+2026-09-20+47973f373a20, window 2026-05-10 to 2026-09-20, read live 2026-09-20 at https://machinerelations.ai/index and the linked domain profiles. ↩