Perplexity answers every question by searching the web "in real-time", using its own index of more than 200 billion URLs, and it shows numbered citations for the sources it used. It ranks short passages, not only whole pages, and prefers documents from "authoritative domains".
To get cited: allow PerplexityBot in robots.txt and in your firewall, write pages where each section answers one question on its own, and be listed on the sites Perplexity already cites for your topic. You cannot pay for a place in the answer. Perplexity says its ads never influence answers and its source labels are not affected by payments.
How does Perplexity choose its sources?
Perplexity has published more about its search engine than most AI companies. In a September 2025 engineering post about its search system, it described the steps:
- Its own index. "Our search index tracks over 200 billion unique URLs." Its earliest version used outside search APIs; now it runs its own crawler, PerplexityBot, plus "third-party crawlers" that it says it requires to respect robots.txt.
- A preference for some sites. To decide which pages to keep ready, it uses models and rules that prioritise "documents from authoritative domains, documents involving undercovered topics" and other pages that enrich the index.
- Passages, not just pages. Pages are split into "self-contained spans, each of which can be individually retrieved and ranked". A clear paragraph deep in a page can be the part that gets cited.
- Two kinds of matching. It searches by keywords and by meaning at the same time and merges the results.
- Several rounds of ranking. Fast scorers narrow the list, then "more powerful cross-encoder reranker models" make the final choice.
What this means in plain English: Perplexity rewards pages where each section answers one specific question in words a buyer would use, on a site it already trusts. A long page that rambles before it gets to the point gives it less to work with than a page with clear headings, each followed by a direct answer.
Quick search, Pro Search and Research
| Mode | What Perplexity says it does | What it means for you |
|---|---|---|
| Standard search | "Surface-level", one or two sources | Only the top one or two pages for the question get cited. You need to be the best match |
| Pro Search | "Comprehensive, dozens of sources" | More room for specialist and local pages next to the big sites |
| Research | "Performs dozens of searches, reads hundreds of sources" | Every narrow sub-question is a separate chance to be read |
The pattern is the same as Google's "query fan-out": a bigger answer is built from many smaller searches. A business that has one clear page for each real buyer question shows up in more of those searches. Our AI Mode guide shows how to map the sub-questions behind a buyer's question.
Source labels: Government, Academic and Trusted
Some citations in Perplexity now carry a small shield icon with one of three labels. Perplexity says it "rates the whole website a source comes from, not each individual page", using checks such as "Does the site correct its mistakes?" and whether it says who wrote each piece. It also says "partnerships, payments, and other business arrangements do not affect a site's label", and that "most domains on the web do not have a label", which is not a negative judgment.
A small business will rarely carry a label, and doesn't need one to be cited. But the checks are a useful list of what Perplexity values in any source: named authors, dates, corrections when you get something wrong, and staying within your own area of expertise.
PerplexityBot, Perplexity-User and robots.txt
| Bot | Job | robots.txt |
|---|---|---|
| PerplexityBot | Builds the search index. "Not used to crawl content for AI foundation models" | Respected. Perplexity recommends allowing it "to ensure your site appears in search results" |
| Perplexity-User | Opens a page when a user's question needs it, and links to it in the answer | "Generally ignores robots.txt rules", because a user asked |
Two details matter. Blocking PerplexityBot does not make you fully invisible: Perplexity says it "may still index the domain, headline, and a brief factual summary" of a blocked page. And Perplexity says it "does not build foundation models", so allowing its crawler does not put your content into model training. For most businesses that makes PerplexityBot an easy yes.
Changes can take "up to 24 hours" to apply. Perplexity publishes its crawler IP addresses, so check that your firewall lets them through. Cloudflare, for example, has asked every new domain since 1 July 2025 whether to allow AI crawlers. On 4 August 2025 Cloudflare also accused Perplexity of using undeclared crawlers to get around blocks, and Perplexity replied that Cloudflare had confused its traffic with another company's. Our AI crawlers list covers that dispute and every other major bot.
What does Perplexity cite for "who should I use?" questions?
We have not run our 500-question study on Perplexity, so we won't claim numbers for it. What we can share is what another live-search assistant, Claude, cited for the same kind of questions, because the mechanics are similar: search, read, cite. Of the 4,634 sources Claude cited for 500 real buyer questions, 79.4% were businesses' own sites and other sites, 9.5% directories and listings, and 38% had a "best" or "top" style title. The mix changed a lot by industry: restaurants were built from TripAdvisor and lists, while home services and car repairs drew mostly on the businesses' own pages.
Perplexity's own description points the same way: "authoritative domains" for the index, and passages that answer the question for the final pick. So the work is the same: one clear page per buyer question on your own site, and a complete, accurate profile on the directories and lists that already rank for your topic.
Shopping, bookings and local results
- Products. Perplexity's Merchant Program is "free for merchants", and members get "increased chances of being a 'recommended product'" because their products are in its index. Its Instant Buy help adds that it is "more likely to recommend merchants who provide deeper product details such as availability, reviews, pricing, and specifications", and that "advertisers cannot pay to list their products". Checkout inside Perplexity is for US users.
- Restaurants. Since 4 August 2025 Perplexity can book tables through OpenTable, choosing among "OpenTable's 60k global restaurant partners". For a restaurant in a market OpenTable covers, a complete OpenTable listing is part of being recommended.
- Local services. Perplexity publishes no guide for local businesses. In practice, it reads the same things other assistants read: your own service and location pages, directory profiles and review sites. Keep your name, address, phone, hours and prices identical on all of them.
What you can and can't control
You can control
- Whether PerplexityBot can reach your site, through robots.txt and the firewall
- Whether each page answers one question clearly, near the top
- Named authors, dates and corrections on your content
- Your product data, if you sell online
- Your profiles on the directories and lists Perplexity cites
You can't control
- Which domains Perplexity treats as authoritative
- Which mode the user picks, and so how many sources are read
- Source labels, which Perplexity assigns itself
- Answers changing between runs and over time
- Buying your way in: Perplexity says ads do not influence answers
Checklist: getting cited by Perplexity
- Allow PerplexityBot in robots.txt, and let its published IP ranges through your firewall.
- Ask Perplexity your top 10 buyer questions, in each language you sell in, and save the numbered sources.
- Open those sources. They are the sites you need to be on, or the pages you need to beat.
- Write one page per real question. Put a direct answer under each heading, so every section stands on its own.
- Add the author's name, a date and your sources to guides and articles. Fix mistakes openly.
- Put prices, hours, areas and specifications in plain text, not only in images or PDFs.
- If you sell products, join the Merchant Program and keep availability, prices and reviews complete.
- If you run a restaurant, complete your OpenTable listing where it operates.
- Keep your facts identical across your site, directories and review profiles.
- Re-check the same questions every month and note which sources come and go.
What we'd do for you
We start with a free AI check: 10 real buyer questions and the businesses AI names instead of you. The AI visibility audit (AED 1,950 in the UAE and GCC, USD 549 elsewhere at adjoltz.com/us) covers 30 buyer questions per language on four AI tools, with every cited source saved and the sites AI trusts in your industry listed. Monthly plans (Foundation AED 4,500 or USD 1,249, Growth AED 8,500 or USD 2,349, Authority AED 16,000 or USD 4,399) track your questions on ChatGPT and Gemini (Google AI), the two largest AI referral sources, and the same pages and listings serve Perplexity. Month to month, no setup fee, and an unconditional 30-day money-back guarantee (rules). We work with businesses worldwide.
What we don't promise
- A Perplexity source label for your site. Perplexity assigns those itself.
- That you will be cited for every question, or the same way each time.
- Numbers for Perplexity that we have not measured. When we publish Perplexity data, it will be on our research pages with the method.
- Any trick that would count as manipulation. Perplexity has said it wants to avoid "arbitrary tactics" that improve rankings "at the expense of user utility".