How I Use the Citation Leaderboard

I use one editorial leaderboard when I decide where a story should live, and I only use it if every row shares a denominator. On that board Medium ranks first, Forbes is cited four times as often as TechCrunch, and Reuters does not yet clear the evidence floor. That is not the prestige order I was handed for most of my career. It is the order of selection on a fixed set of answer runs.

The publication study is Which Publications AI Search Engines Cite Most in 2026. That page is the deeper dataset. This essay is how I would use it as a founder: which column I read before I pitch, what I do with a wire that looks busy, and the release that would send me back to a count of URLs.

The board I am willing to put in a meeting

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. A rate or a rank waits until the domain clears 10 observed runs across 7 distinct run dates.

Publication	Citation rate	Cited runs	Days cited	Engines	Confidence	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

Forbes is cited in 665 runs against TechCrunch's 164. Four times, on the same 16,039. Medium, which a lot of pitch decks file under open publishing and then set aside, is the most-cited editorial publication on this denominator, on 109 of 127 days, across all six engines. Reuters at 13 runs is below the floor. I leave it unranked rather than invent a rank the release refused to publish.

The editorial-class rank needs the classifier in the room. source_role in this release mixes source-type evidence with AuthorityTech's placement catalog. 96 of 1,231 class members enter through the catalog, and the Index is moving that classification onto source-type evidence alone. Six platform domains sit in the class with the newsrooms: medium.com at rank 1, amazon.com at 7, prnewswire.com at 8, apple.com at 16, wordpress.com at 239, blogspot.com at 330. Substack, beehiiv, and dev.to are filed as community platforms. Those six hold 1,358 of the class's 13,525 cited runs, 10.04%, including the top slot. The class denominator has been published from 1,025 to 1,230 domains because each release counts who was classified. The rate, the runs, the days, the engines, and the confidence grade are observations on a fixed run set. When the rank and the rate disagree with my instinct, I follow the rate.

What a count does to the same names

A leaderboard is only as good as its denominator. A counter that only climbs is counting from the day it was switched on, and what it counts is URLs. A newsroom that publishes more URLs outranks a newsroom that publishes fewer, regardless of which one an engine selects. The Index counts answer runs over a window. That is why Forbes sits above TechCrunch here. I make the same mistake in my own week when I treat a full calendar as proof that the hard decision got made. Volume feels like momentum. Selection is the thing the buyer meets when they ask ChatGPT, Gemini, Claude, or Perplexity a real question.

Syndication still has a job. It is not this job. The wire and press-release 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. The Index files prnewswire.com as an editorial publication, so that nine-domain class leaves it out. Counted with the wires, the class is ten domains and 414 of 112,518 cited runs, 0.37%. It still holds, because prnewswire.com's best standing in any published segment is 15 of 448, and the class still holds none of the 880 top-ten slots. A press-release page can be structured and crawlable. Crawlable is not chosen.

Conductor's 2026 AEO/GEO benchmarks report made the macro claim I keep next to this table: AI is replacing the website as the first place many buyers encounter a brand. Muck Rack's Generative Pulse reporting, as I use it here, puts the same pressure from the other side. More than four-fifths of tracked AI citations came from earned media, and journalism becomes more dominant when the query implies freshness. I do not add those studies to the 16,039 and pretend I have a larger denominator. They are a second witness. The Index is the one I will re-pull by release id.

Syndication expands the surface. An editorial selection changes the answer. That split is why distribution, entity clarity, citation architecture, and measurement are different layers in the Machine Relations stack, and why I keep the founder's version of the order on Machine Relations. One dashboard that declares a winner by adding every URL together is emotionally easier. It is also how a wire beats Forbes in a meeting and loses the answer. The days-cited column is the one a raw counter cannot fake. Medium on 109 of 127 days and Forbes on 106, both across six engines. TechCrunch on 44 days across five. Business Insider on 41. Fortune on 28. Reuters on 13. The machine is returning to a narrow set, on different questions, on different dates. That recurrence is what volume was standing in for.

A mention that sends a few hundred people to the site and is never reused is a different asset from a placement on a publication the models cite again on a buyer-intent question. One is attention. The other becomes part of the answer later. I want the second when the budget line says citation. I want the first when the budget line says attention. I have stopped letting one report serve both.

How I would use the days-cited column in a pitch meeting

If the room wants one column, I give them days cited, then the rate, and I refuse to let rank travel alone.

Medium at 109 of 127 days and Forbes at 106, both on six engines at confidence A, are sources the release shows being selected again and again. TechCrunch at 44 of 127 days, five engines, confidence B, 164 runs against Forbes's 665, is a real newsroom on a thinner recurrence. Business Insider at 41 days and Fortune at 28 days, confidence C for Fortune at rank 36, are further down the same slope. Reuters at 13 days and 13 runs is under the floor. I would say those sentences before I say "rank 1" or "rank 9," because the class rank still has six platform domains inside it holding 10.04% of the editorial cited runs, and the class size moves between 1,025 and 1,230. The rate on 16,039 runs does not move when the catalog does.

A wire advocate in that meeting gets the class test, not PR Newswire's rank 8. Nine domains, 232 of 112,518, 0.21%. With prnewswire.com counted in, ten domains, 414 runs, 0.37%, and still 0 of 880 top-ten slots, because its best segment standing is 15 of 448. I can hold rank 8 and the empty top-ten list in the same minute. They are different denominators. The pitch that adds them into a single "wires are winning" sentence is the pitch I stop.

The decision

I would choose the outlet for a citation-seeking story from this kind of board: one window, one run count, a rate, a days-cited figure, an engine count, and a confidence grade. On this release I would take Medium or Forbes before TechCrunch when the only goal is selection frequency, and I would say so even if the room expects the reverse. I would leave Reuters off a rank-ordered target list until it clears the floor. I would buy a wire for distribution and disclosure, and I would not enter its pickup count as a forecast of citations.

The scope is editorial choice for answer-engine citation on release mri_score_v2.0+2026-09-20+47973f373a20. It does not say a Fortune story is a bad story for a human reader. Fortune's 59 runs and rank 36 are a citation fact at confidence C. They are not a judgment of the newsroom. It does not say the class rank is sacred. Ten percent of the editorial class's cited runs sit on six platform domains, and the denominator moves. The rate is what I re-check.

I would go back to ranking by URL counts if a later release, on a stated window, showed the rate order and the count order selecting the same outlets for the buyer questions I care about, and showed the wire class taking top-ten slots rather than 0 of 880. I would also reorder this list the week Reuters clears the floor with a rate that belongs in the conversation, or the week Medium's rate falls to the pack while its class rank is exposed as catalog membership rather than selection. Until then, days cited is the column I point at, and a leaderboard without a denominator does not get a pitch written from it.

The human side of the same choice, which outlets carry a story once the third-party surface is real, is something Christian Lehman, Chief Growth Officer, writes about from the publication side. The definitions I want in the meeting, so "rank" and "rate" do not get swapped, are in the glossary. The engines document their own crawlers: Google, OpenAI, Perplexity, and Anthropic. A citation claim should be as specific as those pages. A rate, a denominator, a window, an engine count, a grade.

FAQ

Why is Medium first if I was taught to chase Forbes and TechCrunch?

Because the unit is answer runs, not masthead and not URLs shipped. Medium is 808 of 16,039 runs, 5.04%, 109 of 127 days, six engines, confidence A. Forbes is 665 runs, 4.15%, 106 days, the same engines, the same grade. TechCrunch is 164 runs, 1.02%, 44 days, five engines, confidence B, rank 9. I use that order for citation frequency. I use a different order when the reader is a person who will never open Medium.

Why is Reuters unranked?

Thirteen cited runs, 13 of 127 days, three engines, confidence marked Collecting, below the floor of 10 runs across 7 dates at which this release will publish a rank. I will not promote it into the target list to protect a prior belief about trust. The floor is the point. A famous newsroom can be under-sampled or under-selected. This release does not let me pretend those are the same, and it does not let me rank the row anyway.

Does PR Newswire's rank 8 mean the wire strategy won?

It means prnewswire.com is filed inside the editorial class in this release and is cited in 182 runs, 1.13%, confidence B. The wire class as a class is 0.21% of classified citation without it and 0.37% with it, and it holds 0 of 880 top-ten slots. Rank 8 of a mixed class is not a top-ten slot in a buyer segment. I read both numbers before I move money.

What do I ask a vendor who shows a different order?

The window, the denominator, the evidence floor, and whether the number can fall. A figure that only rises is a counter. I also ask whether platform domains are inside the class, because six of them hold 10.04% of editorial cited runs here, including first place. If the vendor cannot separate rate from class rank, I use the rate and I keep their board out of the decision.