How to Get Cited by Perplexity: Be a Source It Already Trusts
To get cited by Perplexity, become a source it already trusts. Index data from 6,847 Perplexity answers shows which source types it cites and what to fix first.
To get cited by Perplexity, become a source it already trusts for the question, then make the page that carries your answer easy to reach, quote and check. Perplexity searches the web live on every question, so a reachable page with a direct answer can be picked up once. Being cited again and again goes to the sources it leans on across thousands of answers: editorial publications, community platforms, reference databases, research, and the companies whose own sites are the original record for what they sell.
That second half is where most plans stop short. A page-level checklist gets you into the candidate set. Repeat citation goes to sources it already trusts. Ranking in Perplexity means the same thing as getting cited by it: an answer has no position one to hold, only the sources it cites.
Key takeaways
- Answer first, then prove it. Perplexity searches the web for each question, and a page that answers that exact question in its opening is the easiest one to use. Put the answer in the first two sentences and the evidence right after it.
- Let both Perplexity agents in. PerplexityBot surfaces pages in Perplexity search results; Perplexity-User fetches a page when a user's question needs it. A firewall that blocks either one makes the page unreachable.
- Be where Perplexity already looks. In the Machine Relations Index release of October 21, 2026, editorial publications took 30.4% of Perplexity's classified citations, vendor-owned sources 28.4% and community platforms 12.7%. reddit.com alone was cited in 22.1% of the 6,847 Perplexity answer runs observed.
- Being read is not being cited. In a study of about 14,000 real-world LMArena conversations with search-enabled models, Perplexity's Sonar visited approximately 10 relevant pages per query and cited three to four.
- A first citation is easier than a repeat one; repeat citation is the signal. Perplexity cited 16,383 distinct domains in that release, about twice as many as ChatGPT, Claude or Gemini on similar run counts, partly because each Perplexity answer cites more sources. Most of the domains only Perplexity cited were cited once.
- Wire services are a small share of what Perplexity cites. Wire distribution took 0.5% of Perplexity's classified citations, against 4.1% of ChatGPT's.
- This is one engine's slice of Machine Relations, the discipline of earning AI citations and recommendations that Jaxon Parrott coined in 2024. The same work applies to every answer engine, not only Perplexity.
The short answer: five steps to get cited by Perplexity
- Make the page reachable. Allow PerplexityBot in robots.txt, and allow both PerplexityBot and Perplexity-User at the firewall or CDN. Then read your server logs for both agents.
- Answer the exact question first. The opening sentence answers the question in plain words. No preamble, no definition of the industry, no story before the answer.
- Publish evidence only you have. Original measurement, first-party data, a documented method. A summary of someone else's study gives Perplexity a reason to cite the original instead.
- Earn presence in the sources Perplexity cites. Accurate coverage in editorial publications, useful answers in the community threads it reads, complete profiles in the databases it trusts. This is where Perplexity's repeat citations go in the Index, so it is where your claim needs to be.
- Measure per engine and over time. Track whether Perplexity cites you again across many runs, not whether it cited you once. A single citation is an observation, not a rate.
What Perplexity cites, from 6,847 observed answers
The Machine Relations Index runs a fixed library of buyer questions every day across six answer engines and records the root domain of every source each answer cites. Perplexity is observed through Perplexity's Sonar API, the Index's stand-in for the Perplexity app. In the October 21, 2026 Machine Relations Index release (every category and question shape, May 10 to October 21, 2026), Perplexity was observed on 6,847 answer runs.
Each cited domain carries a source role. Counting one citation per domain per answer run, 26,496 of Perplexity's 80,756 citations went to domains with a classified role; the rest went to domains the Index lists as other observed sources. Among the classified citations:
| Source role | Share of Perplexity's classified citations | Share of ChatGPT's classified citations |
|---|---|---|
| Editorial publication | 30.4% | 41.9% |
| Vendor-owned source | 28.4% | 22.2% |
| Community and social platform | 12.7% | 4.0% |
| Academic and government source | 11.5% | 11.1% |
| Market and company database | 7.2% | 5.4% |
| Analyst and consulting research | 6.6% | 4.4% |
| Search or media platform | 2.8% | 6.9% |
| Wire and press-release distribution | 0.5% | 4.1% |
ChatGPT here is the Index's API observation of it on the same questions, 6,735 runs in the same release. The ten domains Perplexity cited most often, by share of its 6,847 runs:
| Domain | Source role | Perplexity runs citing it |
|---|---|---|
| reddit.com | Community and social platform | 22.1% (1,511) |
| linkedin.com | Community and social platform | 11.6% (794) |
| youtube.com | Search or media platform | 9.7% (664) |
| forbes.com | Editorial publication | 8.2% (559) |
| g2.com | Market and company database | 5.2% (359) |
| nih.gov | Academic and government source | 4.7% (324) |
| arxiv.org | Academic and government source | 4.1% (279) |
| nytimes.com | Editorial publication | 4.0% (276) |
| yahoo.com | Editorial publication | 3.9% (269) |
| amazon.com | Other observed source | 3.8% (262) |
Method note. Shares are computed from the public Machine Relations Index dataset, the October 21, 2026 Machine Relations Index release: for each domain, the Perplexity entry in its per-engine observations gives the runs that cited it, summed by the domain's published source role. Each domain page shows the same per-engine figures; the Index API serves per-domain totals and the engines citing each domain. The Index records observed citation, not why an engine chose a source.
What the source mix means for your plan
Editorial publications are the largest trusted pool. Nearly a third of Perplexity's classified citations go to editorial publications such as forbes.com, nytimes.com and yahoo.com. A claim that lives only on your own domain competes with every outlet Perplexity already cites on the topic. Being described accurately in those outlets puts your claim where Perplexity already looks. Which outlets AI engines cite in each buyer category is on AuthorityTech's publication pages, built from the same Index.
Community platforms matter more on Perplexity than on ChatGPT. Community and social platforms take 12.7% of Perplexity's classified citations and 4.0% of ChatGPT's. reddit.com is the most-cited domain in the whole Index, and in this release its citations come from Perplexity, Gemini, Google AI Mode and Google AI Overviews, not from ChatGPT or Claude. The threads where buyers ask questions are a source Perplexity cites and, in this release, ChatGPT and Claude do not. How to earn those citations is below.
Owned pages do get cited, when they are the primary source. Vendor-owned sources take 28.4% of Perplexity's classified citations; microsoft.com, hubspot.com, zapier.com and ibm.com are among the 25 domains it cites most. These are companies whose own sites are the original record for their products. That is the bar for your own pages: be the primary source for what you describe.
Research and databases carry the facts. Academic and government sources (nih.gov, arxiv.org) and market databases (g2.com) together take 18.7% of Perplexity's classified citations. If buyers compare you on a database Perplexity reads, an incomplete profile there is a gap in your answer.
Wire distribution is a small share of Perplexity's citations. Wire and press-release services take 0.5% of Perplexity's classified citations and 4.1% of ChatGPT's. prnewswire.com, for one, was cited in 12 Perplexity runs and 135 ChatGPT runs. A release can still put a fact on record; it is a small part of what Perplexity cites.
How to earn Reddit citations in Perplexity
reddit.com is the single domain Perplexity cites most in the Index: 1,511 of its 6,847 answer runs in the October 21 release. That makes Reddit worth doing properly, and doing it properly is not posting about yourself.
- Find the threads that already answer your buyers' questions. Ask Perplexity the questions buyers ask in your category and note which threads it cites. Those threads are already in the candidate set; a new post of your own starts outside it.
- Answer the question in the thread, not your pitch. A useful reply names the trade-off, the number and the condition it holds under: the same answer-first shape that makes your own page quotable.
- Post as a named person and say who you work for. Each subreddit sets its own rules on self-promotion, and moderators remove posts that break them. Read the rules first; where vendors may take part, a disclosed expert answer is within them and an anonymous plug is not.
- Link evidence only where it answers the question. Your primary data, method page or documentation earns its place in a reply. A product page usually does not.
- Say the same thing everywhere else. A thread is one source among the ones Perplexity reads for a question. Your own page, your coverage and your database profiles should carry the same claim the thread does.
Reddit supplements the editorial and database presence above; it does not replace it. In this release reddit.com's citations came from Perplexity, Gemini, Google AI Mode and Google AI Overviews, not from ChatGPT or Claude, so a Reddit-only plan reaches some engines and misses others.
Where B2B software answers come from
For a B2B software company, the sources that settle the answer are the ones buyers already use to compare vendors. In the Machine Relations Index's enterprise software category, the October 26, 2026 Machine Relations Index release, these are the domains all six engines together cited most often across the category's 1,819 answer runs:
| Domain | Source role | Category runs citing it |
|---|---|---|
| erpresearch.com | Market and company database | 12.3% (223) |
| linkedin.com | Community and social platform | 8.6% (157) |
| zapier.com | Vendor-owned source | 5.2% (94) |
| reddit.com | Community and social platform | 4.9% (90) |
| g2.com | Market and company database | 4.3% (79) |
| monday.com | Vendor-owned source | 4.2% (77) |
| zylo.com | Vendor-owned source | 4.2% (76) |
| techradar.com | Editorial publication | 4.1% (75) |
| techtarget.com | Editorial publication | 4.1% (75) |
| atlassian.com | Vendor-owned source | 3.9% (71) |
Method note. These rows are the category's top cited domains in the public Machine Relations Index dataset for the October 26, 2026 Machine Relations Index release, a later release than the Perplexity figures above. The public dataset does not split category figures by engine, so this is the enterprise software category's source pool across all six engines, not Perplexity's alone.
Three moves follow for a B2B brand:
- Complete the databases buyers compare you on. Two of the five most-cited domains in the category are market databases. An empty or outdated profile on the comparison sites in your category is a gap in the answer before any page of yours is read.
- Be the primary record for your own product. Four vendor sites are in the category's top ten, each the original source for how its own product works. Documentation, pricing logic and integration pages that answer a buyer's question are the owned pages engines cite.
- Earn coverage in the technology press your buyers read. The editorial domains in the category's top ten are technology publications, techradar.com and techtarget.com. Which outlets AI engines cite in each buyer category is on AuthorityTech's publication pages.
Being read is not being cited
Perplexity reads more pages than it credits. In The Attribution Crisis in LLM Search Results, researchers analysed about 14,000 real-world LMArena conversations with search-enabled models and found that Perplexity's Sonar visited approximately 10 relevant pages per query but cited only three to four; the average Sonar query left about three relevant websites uncited. The counts depend on the setup: the Index's buyer questions, run through the Sonar API, carried about 11.8 cited domains per answer. The lesson does not: a page Perplexity reads can still go uncited. Reaching the page set it reads is the entry ticket; being the source it credits is a separate contest, and the checks for each stage are in how Perplexity decides which sources to cite.
The width trap: why one citation proves little
Perplexity reads the widest source pool of any engine the Index observes. In the October 21, 2026 Machine Relations Index release:
| Engine | Observed runs | Distinct domains cited |
|---|---|---|
| Perplexity | 6,847 | 16,383 |
| ChatGPT | 6,735 | 9,009 |
| Claude | 6,767 | 8,416 |
| Gemini | 6,225 | 8,033 |
Perplexity is observed through its Sonar API; ChatGPT, Claude and Gemini are the Index's API observations of those engines. Part of the width is per answer: Perplexity's answers carried about 11.8 cited domains per run, against about 4.2 for ChatGPT.
Two more numbers from the same release decide how to read that width:
- Most of it is a one-time tail. 8,132 domains were cited by Perplexity and no other engine. 4,429 of them (54.5%) were cited exactly once in the 158-day window; 644 were cited five times or more.
- The sources it relies on are everyone's sources. 99 of the Index's 100 most-cited domains are cited by Perplexity (the exception is itpro.com), and 66 of the 100 are cited by all six engines.
So the door is wide and most of what walks through it is seen once. A crawlable page with a direct answer has a real chance of one Perplexity citation. Repeat citation, the thing that makes a brand part of the answer, goes to the same cross-engine sources every engine leans on. Machine Relations Research's analysis of Perplexity's source pool tracks the same pattern release by release.
Make the page reachable and quotable
Reachable. Perplexity's crawler documentation names two agents. PerplexityBot is designed to surface and link websites in Perplexity search results, and Perplexity recommends allowing it in robots.txt. Perplexity-User supports user actions: when a user asks a question, it may visit a page to help answer and link it. Allow both at the firewall or CDN as well as in robots.txt, because a block there makes the page unreachable whatever robots.txt says. Perplexity says robots.txt changes can take up to 24 hours to reflect, so check the logs the next day.
Quotable. An answer engine quotes sentences, not pages. Give every section one sentence that carries the section's claim on its own. Use headings that match the questions buyers ask. Put numbers next to their source, date and denominator, so the sentence survives being quoted out of context.
Clear about who you are. The same company name, category and description everywhere you control, and requested everywhere you do not. If one page calls you an AI SEO agency, another a PR firm and your coverage something else again, there is no stable entity for Perplexity to attach a claim to. Measure it with entity resolution rate before rewriting anything.
Current. Update the evidence when it goes stale and keep the modified date honest. Freshness is maintenance on everything above, not a separate tactic.
The 90-day order
- Days 1 to 7: access. Read the server logs for PerplexityBot and Perplexity-User. No requests or repeated blocks is an infrastructure defect, and nothing below it matters until it is fixed.
- Days 7 to 21: answer-first rewrites. Rewrite your highest-intent pages so the opening answers the query and every section survives as one sentence.
- Days 21 to 45: one entity. One company descriptor, one category name, one founder bio, everywhere.
- Days 45 to 75: proof only you have. Benchmarks, method pages, original measurement.
- Days 1 to 90, in parallel: trusted presence. Coverage in the editorial publications, communities and databases Perplexity already cites for your category. This work has the longest lag, so it runs alongside steps 1 to 4 from the start rather than after the owned work disappoints.
How to recover Perplexity citations a competitor took
Check three things in order. First, freshness: is your evidence older than theirs? Second, original proof: did they publish data you do not have? A rewrite of the same claims will not beat new evidence. Third, trusted presence: are they described in the editorial publications, threads and databases Perplexity cites for the question, and you are not? A single page refresh does not change where Perplexity looks; that gap is in the sources it already cites for the question. Never copy the competitor's page. Perplexity can read both, and the original wins.
Where getting cited by Perplexity sits: Machine Relations
Getting cited by Perplexity is one engine's slice of Machine Relations: the discipline of earning AI citations and recommendations for a brand by making that brand legible, retrievable and credible inside AI-driven discovery. Jaxon Parrott coined Machine Relations in 2024. AuthorityTech, which he founded in 2018, is The World's First Machine Relations Agency. Machine Relations is open to anyone, and machinerelations.ai is its home, where Machine Relations Research publishes the Index this page reads. GEO and AEO are tactical layers inside it: page structure and answer formatting are the GEO part of this guide, and being the source an engine trusts is the whole of it.
That is why the five steps above are not Perplexity tricks. Reachable pages, answer-first structure, original proof and presence in trusted sources are what every answer engine reads.
Where to go deeper
This page is the intervention answer: what to do, in what order. Neighboring questions are one hop away:
- The retrieval mechanism itself, stage by stage, is in how AI platforms like Perplexity decide what to cite. That page owns the mechanism; this one owns the response to it.
- Whether a specific page clears each stage Perplexity applies, from page availability to visible attribution, is the diagnostic in how Perplexity decides which sources to cite.
- If you sell B2B SaaS, the segment version is getting cited in Perplexity as a B2B SaaS brand. For three engines at once, read getting cited by ChatGPT, Perplexity and AI Overviews.
Frequently asked questions
How do I get my website cited by Perplexity?
Make the page reachable to PerplexityBot and Perplexity-User, answer the exact question in the first sentences, publish evidence only you have, and earn presence in the sources Perplexity already cites: editorial publications, community platforms, databases and research. The first three make a page a usable candidate; the fourth puts your claim in the sources Perplexity cites again and again.
How do you rank in Perplexity?
Ranking in Perplexity means being one of the sources its answer cites. Make the page reachable to PerplexityBot and Perplexity-User, answer the question in the opening, publish evidence only you have, earn presence in the editorial publications, communities and databases Perplexity already cites, and measure whether it cites you again across many runs.
What kinds of sources does Perplexity cite most?
In Machine Relations Index release of October 21, 2026, editorial publications took 30.4% of Perplexity's classified citations, vendor-owned sources 28.4%, community and social platforms 12.7% and academic and government sources 11.5%. The most-cited single domain was reddit.com, in 22.1% of 6,847 Perplexity answer runs.
Does Perplexity cite more websites than ChatGPT?
Yes. In the same release Perplexity cited 16,383 distinct domains across 6,847 answer runs and ChatGPT 9,009 across 6,735, on the same buyer questions. 8,132 domains were cited by Perplexity alone, and most of those were cited once.
Does Perplexity cite Reddit?
Yes, more than any other domain. reddit.com was cited in 1,511 of 6,847 Perplexity answer runs (22.1%) in the October 21, 2026 Index release. ChatGPT and Claude, observed on the same questions, did not cite it in that release.
Why does Perplexity cite Reddit instead of my website?
Because Reddit is the domain Perplexity cites most: reddit.com was cited in 22.1% of Perplexity's 6,847 answer runs in the October 21, 2026 Index release. Threads carry direct answers in buyers' own words. To compete, answer in the relevant threads as a named, disclosed expert, and make your own page the primary record for what those threads discuss.
Do press releases get cited by Perplexity?
Rarely, relative to other sources. Wire and press-release distribution took 0.5% of Perplexity's classified citations in the release, against 4.1% of ChatGPT's. Perplexity does cite the large wires, so a release can put a fact on record, but wires are a small share of the sources Perplexity cites again and again.
Can you pay to get cited by Perplexity?
Perplexity has introduced advertising products. The Index measures organic citation in answers only; paid placement is outside what it records.
How long does it take to get cited by Perplexity?
There is no fixed timeline. Perplexity searches live, so a reachable page with a direct answer can be cited soon after it is crawled. Repeat citation in the Index goes to editorial publications, communities and databases, and being present there takes longer to build, which is why that work runs from the start.
Is getting cited by Perplexity part of GEO or Machine Relations?
Both, at different levels. Page structure and answer formatting are GEO tactics. Becoming the source an engine trusts, across engines, is Machine Relations, the discipline Jaxon Parrott coined in 2024; AuthorityTech is The World's First Machine Relations Agency.