Short answer. Nobody outside Perplexity can tell you with proof why one page gets cited and another does not. The official developer documentation says few things: PerplexityBot exists to surface and link sites in search results, and Perplexity-User visits pages when a person asks a question. One prerequisite follows that can be demonstrated: do not block PerplexityBot. Everything else, from text structure to freshness, circulates as "technique" but is at best a reasonable conjecture. This article separates what is documented from what is plausible and from what is merely repeated, and gives you a way to measure what can be measured without promising a result.
Verified on 30 October 2026
In the developer documentation I read I found no official source-selection criterion, nor a dashboard for publishers; this does not prove they do not exist elsewhere. The claims about "citation signals" you find on industry blogs are, in the results I read, inferences by marketing practitioners and contradict each other: one source says a blocked site may appear as a domain with a short summary, another that it cannot be cited at all; one estimate speaks of 3-8 numbered citations per answer, another of an average of about 22 across more than 118,000 analyzed answers. I do not use them as facts.
Three columns: documented, plausible, repeated
| Documented by the vendor | Plausible but not proven | Repeated by third parties without evidence I could verify |
|---|---|---|
| PerplexityBot surfaces and links sites in results; it honors robots.txt and allowing it is recommended. Perplexity-User visits pages at a user's request and generally ignores robots.txt. robots.txt changes: up to 24 hours. IP ranges published in two JSON files. | Clear text, answer up front, visible dates, complete HTML not depending on JavaScript, stable URLs, sitemaps with real dates, explicit author identity. | Weights of the "signals" (authority, freshness, structure), preference for "community" sources, ideal paragraph length, average number of citations per answer, whether or not the crawler executes JavaScript. |
The middle column is not a list of tricks: it is a list of things that improve the site for readers and for any crawler. The criterion for including them is that if the conjecture is wrong, you lost nothing: putting an answer in the first lines of a page, or showing the update date, serves whoever reads anyway. Practices that cost a lot for a hypothetical benefit (massive rewrites, pages made for engines) I do not recommend.
What can really be measured
Four measures, each with its limit. None tells you "how often you are cited".
- Crawler visits (server logs). How many PerplexityBot requests per URL, with addresses verified against the official file (article 7). Limit: being visited does not mean being cited.
- Visits from perplexity.ai (analytics). People who clicked a source. Limit: undercount, because not every citation is clicked and not every browser passes the referrer.
- A probe with the Search API. Repeatable and documented (article 3): your questions, filtered to your domain, recording whether your URLs appear among the results. Important limit: it measures whether your pages are retrievable by the Search API; the documentation does not say the index and ranking are identical to those of the consumer application, so it is not a measure of the app's citations.
- Manual tests. Asking the same questions in the application. Limit: answers vary between attempts and may depend on person, place and time. A single appearance proves nothing, nor does it count as a trend.
import datetime as dt, json
from perplexity import Perplexity
client = Perplexity()
DOMAIN = "example.com" # your domain
QUESTIONS = [ # 20 questions a reader of yours would really ask
"how is withholding tax calculated for a freelancer",
# ...
]
def probe(path="probe.jsonl"):
today = dt.date.today().isoformat()
with open(path, "a", encoding="utf-8") as f:
for q in QUESTIONS:
res = client.search.create(
query=[q],
max_results=10,
country="US",
search_domain_filter=[DOMAIN], # allowlist
).results
f.write(json.dumps({
"date": today,
"question": q,
"n_own_urls": len(res),
"urls": [r.url for r in res],
}, ensure_ascii=False) + "\n")
probe()
Cost: 20 questions a week are about 80 requests a month: 0.40 dollars at 5.00 per thousand. Mind what the domain filter changes: this way you ask "among this site's pages, which are retrievable for this question?", not "does my site beat the others?". For the second question you would search without the filter and look at your position, but again in the Search API, not in the app. With 20 questions the noise is large: treat only wide, lasting changes as significant, not a comparison between two weeks.
What I choose not to write
- Percentage improvements in citation after this or that change: I found no independent, reproducible measurements.
- "Perplexity does not execute JavaScript" as a fact: I met it as a repeated claim, not in official documentation. Publishing complete server-side HTML remains good practice regardless, but I do not present it as a cause of citations.
- A "rank on Perplexity" checklist: if it is sold as a guarantee, it is selling you a conjecture.
- This site's data. The site allows PerplexityBot in
robots.txt, but I publish no citation numbers because I have not measured them with a method that would withstand criticism.
Use cases: citation as a channel, by sector
Hypothetical scenarios; declared assumptions, no measured data.
Small business, professional practice: the law firm and citation as authority
Who: a four-lawyer firm with an insights blog. What it did: one article a month, promoted on social media and by hand in a newsletter. What changes: allow PerplexityBot, put publication and last-verified dates at the top, state the scope (which rule and which date the text refers to), and run the monthly probe. The expected value is not traffic: it is that someone receiving a synthesized answer with a source sees the firm's name. What goes wrong, and here it costs: a rule changes and the article stays cited in its old version; a reader acts on an answer attributing to the firm a claim the article did not make. Remedy: scheduled review, a notice at the top of aging pages, and the rule that no blog article replaces an opinion. Whether being cited brings clients, I do not know and do not promise.
Small business, retail: citation as product discovery
Who: an online mountain-gear shop. What changes: product pages with specifications in text (materials, weight, intended use) and a buying guide per category; the value is in an answer naming the right product for the right question ("what do I need for a three-day trek"). What goes wrong: prices and availability cited out of date; the answer cites a page for a product you no longer sell. Remedy: sold-out pages that clearly say "no longer available" and point to the alternative, not deleted pages. The comparison that matters is between visits from perplexity.ai and conversions, measured over time, not the number of citations (which you cannot see).
Large enterprise, healthcare: health information and responsibility
Who: a healthcare facility with information pages on tests and preparation. What matters: these are health contents, where a citation out of context can do harm. What changes: pages with declared clinical review, date, scope ("for adults", "absent condition X"), contacts to ask staff, and no dose or instruction without the context that makes it safe. What goes wrong: the answer reports a sentence detached from its warning; the reader ignores the rest. You cannot control how an engine clips text: you can only write pages where each important sentence carries its condition with it, and periodically check, with people, what the answers citing the facility say. The control is a review process, not a configuration.
Daily life: the neighborhood association
Who: a village's local association with a site made by a volunteer. What it did: publish the festival programme as an image of a flyer. What changes: a page with the programme written as text, date and place up front. Text is readable also by someone asking an assistant "what is on in the village this weekend". What goes wrong: the date changes and the answer, for days, reports the old one. The volunteer must know this cost: every change to the programme goes on the page, not only on social media. And anyone looking up an event for practical reasons (times, parking) should open the cited source instead of trusting the synthesis.
← Article 7: PerplexityBot and robots.txt Back to article 1: the map →
Frequently asked questions
How do I get cited by Perplexity?
The only documented prerequisite is not blocking PerplexityBot. For source-selection criteria I found no official documentation: practices like up-front answers and visible dates are reasonable but not proven.
Can I measure how often I am cited?
Not reliably from outside. You can measure crawler visits from logs, clicks from perplexity.ai in analytics and your pages' retrievability with the Search API, but none of these is the number of citations in the application.
Are there confirmed ranking signals for Perplexity?
Not in the developer documentation I read. Industry blogs list some, but they are inferences and partly contradict each other.
Does being cited bring traffic?
It is not proven in general. For your site you can compare visits from perplexity.ai with conversions over time, bearing in mind that the measure undercounts citations that are not clicked.
Sources and further reading
- Perplexity, Bots (read on 30 October 2026)
- Perplexity, Search API quickstart (30 October 2026)
- Perplexity, Pricing (30 October 2026)
- Industry blogs on "citation signals": summary of the results of a web search on 30 October 2026, not read in full and not listed one by one because their claims are unverifiable; cited only to say that they contradict each other.
- Cloudflare, 4 August 2025
Commenti
Caricamento commenti...
Sign in to comment