How Perplexity chooses sources: what marketers can influence
How Perplexity retrieves, ranks and cites sources, what it means for your pages, and the levers marketers can actually pull to get cited.
Perplexity answers by searching its own index, retrieving the passages most relevant to each part of the question, and citing the pages those passages came from. It ranks at the level of sections within pages, not just whole pages. A clear, current, self-contained answer to a specific sub-question can win even on a smaller site. You can't buy placement. What you can influence is crawl access, how answerable and fresh your pages are, and how often trusted third parties mention you.
- Perplexity uses its own index (more than 200 billion URLs, according to its Search API post) and ranks passages within pages.
- Questions are broken into sub-queries, and each one can cite different sources.
- Levers you control: access for PerplexityBot, answer-first sections, freshness, evidence, and third-party coverage.
- Levers you don't control: the model, user context, the source selector and day-to-day answer variation.
Checked 5 November 2026 (CET) against the Perplexity Help Center, Perplexity blog, Perplexity API docs and the outlets cited in Sources. Perplexity renames and moves features often, so confirm labels and limits in your own account. "Our view" marks MentionX opinion. Ledgerline is a fictional company used for illustration.
01What happens between a question and a citation?
A simplified view, based on Perplexity's public descriptions of its search stack:
| Step | What happens | What it means for you |
|---|---|---|
| 1. Crawl and index | PerplexityBot crawls and indexes pages | Blocked pages can't be retrieved (lesson 14) |
| 2. Understand the question | The question is interpreted and often split into sub-queries | Each sub-question is a separate chance to be cited |
| 3. Retrieve | Hybrid lexical (keyword) and semantic (meaning) retrieval | Use the words buyers use, and cover the concept clearly |
| 4. Rank | Multi-stage ranking, including sub-document spans | A strong section can beat a whole page |
| 5. Fetch live, if needed | Perplexity-User may fetch a page for a user request | Live facts, like pricing, should be in HTML |
| 6. Generate and cite | The model writes an answer citing the sources used | Quotable, self-contained sentences are easier to use |
Perplexity's Search API post describes the hybrid retrieval, multi-stage ranking and span-level ranking. The consumer product adds its own layers, such as models, modes, personalisation and source selectors, which aren't fully documented.
02Which signals appear to matter?
No engine publishes its ranking weights. Here's what Perplexity's documentation, public statements and repeatable observation support. Treat the "observed" rows as hypotheses to test.
| Signal | Basis | Practical action |
|---|---|---|
| Crawl access | Documented (crawlers page) | Allow PerplexityBot and Perplexity-User at robots.txt and the WAF |
| Relevance of a passage to the sub-question | Documented (span ranking) | Answer each question in its own section |
| Freshness for time-sensitive queries | Observed, and supported by recency options | Visible "updated" dates, current facts |
| Clear structure | Observed | Question headings, short paragraphs, tables |
| Third-party corroboration | Observed | Reviews, publisher coverage, community threads |
| Original data or primary sources | Observed | Publish your own numbers with methodology |
| Domain reputation | Observed | Long-term brand and link building |
Our view: the biggest gains usually come from the second row. Most pages bury the answer to a specific question inside a long introduction. Fix the structure before chasing anything else.
03Why do different users see different sources?
- Mode. Search, Pro Search, Research and Model Council search differently (lesson 3).
- Source selector. Users can choose Web, Academic, Social or Finance sources, which changes the pool completely.
- Model. Paid users can pick models, which may use sources differently.
- Context. Follow-up questions, Projects with files and connectors, and location all change results.
- Variation. The same question asked twice can cite different sources.
That's why tracking uses a fixed prompt set, repeated runs and shares rather than single screenshots (lesson 19).
04How do you diagnose why a competitor is cited instead of you?
Run this for one prompt where you lose:
- Ask the question in Pro Search. Note every citation.
- Ask: "Which specific claim did you take from each source?"
- Open the competitor's cited page. Find the exact passage.
- Find the closest section on your site. Compare them:
| Check | Competitor passage | Your passage |
|---|---|---|
| Answers the sub-question in the first two sentences? | ||
| Uses the buyer's words? | ||
| Specific (numbers, names, steps)? | ||
| Dated and current? | ||
| Self-contained (makes sense quoted alone)? | ||
| Crawlable HTML (not an image, PDF or tab)? |
- Fix the gaps and recheck in two to four weeks.
05What role do third-party sources play?
For many commercial questions, Perplexity cites reviews, comparison articles, publisher coverage, forums and documentation more than vendor pages. If your brand is missing from those sources, your own site has to carry the whole answer. See the citation gap and earned media for how to find and close those gaps, and Perplexity's source preferences in the engine guide.
06What can't you influence?
- Which model the user picks.
- Whether they use Social or Academic sources.
- Their follow-up questions and Project context.
- Day-to-day variation.
Spend your effort on the levers you control and measure over time.
07Worked example: Ledgerline
Ledgerline is a fictional accounts-payable software company. Details are illustrative.
For "best AP automation for mid-market", Perplexity cited Competitor A's comparison page, a review site and a publisher roundup. Ledgerline's product page wasn't cited.
| Check | Competitor A passage | Ledgerline product page |
|---|---|---|
| Answers in two sentences | Yes: "For mid-market teams, the key criteria are..." | No: opens with a slogan |
| Buyer's words | "mid-market", "approval workflows" | "finance transformation platform" |
| Specific | Criteria table with five rows | Feature list, no criteria |
| Dated | "Updated October 2026" | No date |
| Self-contained | Yes | Depends on the hero image |
Walkthrough
- Ran the diagnosis on five losing prompts. The patterns were the same.
- Added an answer-first "Who is Ledgerline for?" section with criteria and a table.
- Added visible updated dates.
- Moved pricing facts from an image into HTML.
- Briefed PR on the publisher roundup gap.
- Rechecked after four weeks. Ledgerline was cited on two of the five prompts in most runs.
08Common mistakes
- Optimising the whole page instead of sections. Perplexity ranks passages.
- Chasing one screenshot. Track shares over repeated runs.
- Ignoring third parties. Your site is only part of the source pool.
- Assuming paid placement exists. It doesn't.
09Try it yourself
Pick one prompt where a competitor is cited and you aren't. Run the six-row comparison table on the cited passage and your closest section. Write down three specific changes.
10Quick quiz
1. At what level does Perplexity's Search API rank content?
Answer: Sub-document spans (sections within pages), as well as whole pages.
2. Name two things that change which sources a user sees.
Answer: Any two of mode, source selector, model, context and normal variation.
3. Can you pay for placement in Perplexity answers?
Answer: No.
11What should you check before the next lesson?
- One losing prompt has been diagnosed with the comparison table.
- You know which pages are your best candidate passages.
- PerplexityBot access is on your checklist for next lesson.
Previous: Automate Perplexity research with n8n. Next: PerplexityBot, robots.txt and WAF settings. Or return to the Perplexity for Marketers track. Related: How to rank in Perplexity, Why Perplexity results differ, Content chunking for LLM retrieval.
12Common questions
Does Perplexity use Google results? Perplexity runs its own search index and ranking. It has described an index of more than 200 billion URLs with hybrid lexical and semantic retrieval. Classic SEO signals still help, because crawlable, well-linked, authoritative pages tend to do well in any index.
Can I pay to be cited by Perplexity? No. There's no paid placement in answers. Perplexity also stopped the sponsored follow-up question ads it tested from 2024.
Why does Perplexity cite a competitor's blog instead of our product page? Usually because that page answers the specific sub-question more directly, more recently or with clearer evidence. Compare the cited passage with your page section by section.
Questions people ask
3 answersDoes Perplexity use Google results?
+
Perplexity runs its own search index and ranking. It has described an index of more than 200 billion URLs with hybrid lexical and semantic retrieval. Classic SEO signals still help because crawlable, well-linked, authoritative pages tend to do well in any index.
Can I pay to be cited by Perplexity?
+
No. There is no paid placement in answers. Perplexity also stopped the sponsored follow-up question ads it tested from 2024.
Why does Perplexity cite a competitor's blog instead of our product page?
+
Usually because that page answers the specific sub-question more directly, more recently, or with clearer evidence. Compare the cited passage with your page section by section.
Sources
- Perplexity blog, Architecting and Evaluating an AI-First Search APIperplexity.ai
- Perplexity API docs, Perplexity Crawlersdocs.perplexity.ai
- Perplexity Help Center, How does Perplexity work?perplexity.ai
- Search Engine Land, Perplexity stops testing advertising (18 Feb 2026)searchengineland.com
- All accessed 5 Nov 2026.