How I Read Perplexity's Source List

When someone asks me which publications Perplexity cites for a B2B question, I look for engine breadth before I look for a Perplexity-only ranking. On the current Machine Relations Index release, every one of the eight publications below is observed on Perplexity, and Medium and Forbes, the two most-cited editorial publications in the release, are cited on all six engines. Perplexity is not keeping a private list of the leaders. It is one of the engines returning to them.

The publication measurement I am reading 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: the check I run before I believe a "Perplexity prefers X" claim, the wire comparison I keep in the same folder, and the missing per-engine rate that would change how sharp I am willing to be.

The table, and the sentence under it

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. Nothing publishes a rate or a rank until it clears 10 observed runs across 7 distinct dates. Engine breadth here is the list of engines on which the domain was observed at all. It is an observed-engine list, not a per-engine rate.

Publication	Citation rate	Cited runs	Days cited	Engines observed	Perplexity among them
medium.com	5.04%	808	109 of 127	6	yes
forbes.com	4.15%	665	106 of 127	6	yes
prnewswire.com	1.13%	182	75 of 127	6	yes
techcrunch.com	1.02%	164	44 of 127	5	yes
cio.com	0.42%	67	30 of 127	6	yes
businessinsider.com	0.62%	100	41 of 127	5	yes
thenextweb.com	0.12%	20	14 of 127	5	yes
techbullion.com	0.07%	12	11 of 127	3	yes

Eight of eight. Medium and Forbes are the two most-cited editorial publications in the release, on all six engines, on 109 and 106 of 127 days.

I read the editorial-class standing with the classifier note attached. The source_role field in this release combines source-type evidence with AuthorityTech's placement catalog. 96 of the 1,231 class members are admitted through the catalog, and 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 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 moves between releases, with published figures from 1,025 to 1,230, because it counts the domains classified in each release. Citation rate, cited runs, days cited, engine breadth, and confidence are the observations. I read the rate first, and I read "Perplexity: yes" as presence, not as a share of Perplexity's own answers.

Why breadth is the first question

A corpus drawn from one retrieval source makes almost every publication look filtered out by the engines it barely sampled. An absence in that kind of sample is a sampling gap. It is not the engine's editorial judgment. The first question I ask of a per-engine claim is how many engines observed the domain, and on how many runs.

A source cited by one engine is a crawl artifact of that engine's index. A source cited by six engines across 106 separate days is a market position. Medium's 109 days and Forbes's 106 days, both on six engines including Perplexity, are the second kind. TechBullion's 12 cited runs, 11 days, three engines, still including Perplexity, is a different kind of yes. I do not brief those two yeses as the same fact.

The engines publish what they crawl: Google's common crawlers, OpenAI's bots, Perplexity's bots, and Anthropic's crawler policy. Any per-engine claim I am willing to repeat can say how many runs on that engine it rests on, and it refuses to publish below a floor. A number from a system I operate gets the same questions I ask of outside research. Proximity is not verification.

The wire comparison I keep next to the list

I do not let a Perplexity "yes" on PR Newswire become a recommendation to buy the wire. Forbes is cited 3.65 times as often as PR Newswire on this denominator. The wire and press-release class as a whole takes 0.21% of classified citation. The Index files prnewswire.com, the largest wire domain in it, as an editorial publication. Counted with the wires, the class is ten domains at 414 of 112,518 cited runs, 0.37% of classified citation. The direction does not flip. prnewswire.com's best standing in any published segment is 15 of 448, so the class still holds 0 of 880 top-ten slots.

A Muck Rack analysis of more than a million AI citations found 82% came from earned media, with institutional credibility as the sorting variable (Muck Rack). Conductor's 2026 AEO/GEO benchmarks document the shift toward answer engines and source-trust weighting. Both are someone else's studies. I link them because they point the same direction as the table. I do not merge their percentages into the Index rate. The Index does not yet publish a per-engine rate, and a quantified claim about how Perplexity ranks publications against each other needs one. "Observed on Perplexity" is the sentence I will say. "Perplexity's number-one publication is Medium, at 5.04% of Perplexity's answers" is a sentence this release does not support, because 5.04% is the all-engine rate.

The decision

I would build a Perplexity pitch list from sources that are observed on Perplexity and on several other engines, with the days-cited column in view, and I would rank them by the all-engine citation rate while I say out loud that the rate is not Perplexity's private rate. Medium and Forbes clear that bar on this release. A domain observed on Perplexity alone, or on three engines with 11 days, does not get the same budget as a domain on six engines across a hundred days. I would not buy wire distribution because prnewswire.com's row says yes in the Perplexity column. The class share and the 0 of 880 top-ten slots are the citation math. The yes is presence.

The scope is B2B-style publication choice against this release's eight-row read. It does not say Perplexity never cites a niche trade. It says I will not treat a one-engine sample as the engine's taste. The methodology note for the figures is the Index correction record, updated through 2026-09-23.

I would change my mind, and I would start publishing a Perplexity-specific order, on the first release that reports a per-engine citation rate with its own denominator and the same evidence floor of 10 runs across 7 dates. Until that column exists, engine breadth is the first check, and a vendor chart that ranks publications "for Perplexity" without a run count on Perplexity is a chart I send back.

The place I keep this attached to the rest of the work is Machine Relations. The rate, the class, and the evidence floor are named in the glossary. I would rather be slightly blunt about a missing per-engine rate than precise about a number the release does not contain.

FAQ

Does Perplexity cite a different set of leaders?

On this release, no. All eight publications in the table are observed on Perplexity. Medium and Forbes, at 808 and 665 cited runs, are on all six engines. The Index does not yet split the 5.04% and 4.15% rates by engine. I can say Perplexity is among the engines that cite them. I cannot say what share of Perplexity's own answers they hold.

Why is TechBullion a yes if the rate is 0.07%?

Because "yes" here means the domain was observed on Perplexity at least once inside the window: 12 cited runs, 11 of 127 days, three engines, which is enough to clear the floor of 10 runs across 7 dates and not enough to treat it as a pillar. I keep the yes and the rate in the same sentence so the yes cannot be sold as a strategy.

Should I pitch PR Newswire because Perplexity observes it?

I would pitch it for distribution jobs a wire actually does. I would not pitch it as a Perplexity citation strategy. Forbes is cited 3.65 times as often. The wire class, even with PR Newswire counted in, is 414 of 112,518 cited runs, 0.37%, and it holds none of 880 top-ten segment slots. Presence on the engine is a lower bar than being selected into the answer a buyer sees.

What would make a Perplexity-only ranking usable?

A published per-engine rate, a denominator, a window, and the evidence floor. The crawler documentation tells me the engine can see a page. It does not tell me the engine chose it. A sample drawn from one retrieval source will manufacture absences. I ask how many Perplexity runs the claim rests on before I reorder a pitch list around it.