Perplexity draws on a wider pool of sources than any other AI answer engine the Machine Relations Index observes. In release mri_score_v2.1+2026-10-14+7a624b28cd41, Perplexity cited 15,033 distinct domains across 5,812 observed answer runs. ChatGPT (API), Claude and Gemini, observed on a similar number of runs, cited between 7,613 and 8,154. Perplexity cited 48.9% of the 30,757 domains in the Index, and 7,412 of them by Perplexity and no other engine.
That reverses what this page reported in July from a 30-day window, when Perplexity looked like the engine with the narrower pool. The fuller window says the opposite: Perplexity is the widest door among the engines, and the one where getting through it once means least.
How wide is Perplexity's source pool? #
The Index runs the same buyer questions across six engines and records every domain each answer cites. Distinct domains grow with the number of runs, so the fair comparison is between engines observed on similar run counts. Four engines were: Perplexity, Claude, ChatGPT and Gemini, each on 5,308 to 5,812 runs.
| Engine | Observed answer runs | Distinct domains cited |
|---|---|---|
| Perplexity | 5,812 | 15,033 |
| ChatGPT | 5,704 | 8,154 |
| Gemini | 5,308 | 7,835 |
| Claude | 5,805 | 7,613 |
| Google AI Mode | 1,659 | 7,704 |
| Google AI Overviews | 1,035 | 3,009 |
On matched volume, Perplexity cited 1.8x to 2.0x as many distinct domains as each of the other three. Google AI Mode and Google AI Overviews were observed on far fewer runs, so their counts are not comparable on this table; Google AI Mode reached 7,704 domains on 1,659 runs, which marks it as a wide reader too.
Run counts are summed from the release's per-segment engine run counts and total exactly the release's 25,323 runs. Domain counts read each domain's published list of engines that cited it.
Most of the width is a long tail #
Of the 7,412 domains that only Perplexity cited, 4,198 (56.6%) were cited exactly once across the 151-day window. Only 567 were cited five times or more, and the most-cited of them, justdial.com, appears in 42 runs. Others in that group include verywellhealth.com (35), thelines.com (34) and digitain.com (32).
So the door is wide, and most of what walks through it is seen once. A page Perplexity cites is not yet a source any other engine relies on.
Where Perplexity agrees with every other engine #
The head of the Index looks the same from every engine. Of the 100 most-cited domains in the release, 99 were cited by Perplexity; the exception is itpro.com. Sixty-seven of the 100 were cited by all six engines, including linkedin.com, g2.com, forbes.com, crunchbase.com and ycombinator.com.
Two engine-specific gaps hold in this release:
- Reddit is not universal. reddit.com is the most-cited domain in the Index, in 2,429 of 25,323 runs (9.59%), yet its citations come from Perplexity, Gemini, Google AI Mode and Google AI Overviews. ChatGPT and Claude did not cite it in the release.
- Wire distribution reaches Perplexity. prnewswire.com, businesswire.com and globenewswire.com are all cited by Perplexity. The July window had recorded a single Perplexity citation of a wire service; the fuller record does not support calling wire services invisible to it.
The July window also recorded ChatGPT citing LinkedIn zero times. In this release ChatGPT is one of the six engines citing linkedin.com.
What this means for getting cited by Perplexity #
- Entry is the easy part. An engine that cites twice as many domains as its peers on the same questions gives a new or niche page more chances to appear. Most pages that get in are cited once.
- Repeat citation is the signal. The domains Perplexity relies on are the ones every engine relies on: 99 of the Index's 100 most-cited. Being in that head is a cross-engine position, not a Perplexity tactic.
- Measure per engine. A source Perplexity cites may be absent from ChatGPT and Claude, as reddit.com is. A visibility number pooled across engines hides that.
- Read the evidence floor. The Index publishes a rate only after a segment clears 10 observations across 7 run dates. A single Perplexity citation is an observation, not a rate.
What the July analysis showed and what no longer holds #
This page was first published on 2026-07-30 from a 30-day per-engine window of 6,020 domains. It reported that Perplexity cited 1,478 of those domains, read that as a narrower pool, and gave per-engine source-role shares and per-domain Perplexity citation counts. The public release does not carry per-engine citation counts per domain, so those tables cannot be reproduced from it and have been removed. The narrower-pool reading, the wire-service reading and the ChatGPT-LinkedIn reading are each contradicted by the release above.
FAQ #
How many websites does Perplexity cite? #
In Machine Relations Index release mri_score_v2.1+2026-10-14+7a624b28cd41, Perplexity cited 15,033 distinct domains across 5,812 observed answer runs between 2026-05-10 and 2026-10-14, the most of any engine observed.
Does Perplexity use different sources than ChatGPT? #
Partly. On similar run counts Perplexity cited 15,033 distinct domains and ChatGPT 8,154. The most-cited sources overlap: 99 of the Index's 100 most-cited domains are cited by Perplexity and 67 by all six engines. reddit.com is cited by Perplexity and not by ChatGPT.
Is it easier to get cited by Perplexity than by ChatGPT? #
The Index records that Perplexity cites a wider set of domains on comparable volume, and that 7,412 domains were cited by Perplexity alone. Most of those were cited once, so a first Perplexity citation is easier to earn than a repeated one.
Can you pay to rank in Perplexity? #
Perplexity has introduced advertising products. The Index measures organic citation in answers only and records nothing about paid placement.
How Machine Relations measures this #
The Machine Relations Index runs a fixed library of buyer questions across ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity and records the root domain of every source each answer cites. Release mri_score_v2.1+2026-10-14+7a624b28cd41 covers 30,757 cited domains and 25,323 observed answer runs from 2026-05-10 to 2026-10-14. Engine run counts are read from each segment's published question context; the engines that cited a domain are read from its published per-engine observations. The Index reports observed citation, not why an engine chose a source, and it does not publish per-engine citation rates. The public dataset and API carry every figure on this page.