Perplexity chooses sources by running live web searches for your question, pulling the most relevant passages from the pages it finds, and citing the pages those passages came from. To become a source, your page has to be crawlable by Perplexity, rank in the underlying search results, and contain a clear, self-contained answer it can lift and attribute.
That is the short version. Perplexity has never published a full ranking formula, so anyone promising a secret trick is guessing. What we can describe is how the product works from its public documentation and visible behaviour, and which page qualities consistently line up with being cited. It is one piece of the wider picture covered in our guide to AEO and GEO for Google AI Overviews and ChatGPT.
What actually happens when someone asks Perplexity a question?
Perplexity works like a research assistant that searches first and writes second. It doesn't answer only from what a language model memorised during training; it fetches current web pages and builds the answer on top of them.
The process, in plain terms:
- The question gets rewritten into searches. A question like "best CRM for a 10-person real estate team in Sydney" may become several narrower queries about CRM features, pricing and real estate workflows.
- Candidate pages are retrieved. Perplexity draws on its own index (built by its crawler, PerplexityBot) and live search results to collect a shortlist of pages.
- Passages are extracted. Rather than reading whole pages, it pulls the paragraphs, lists and table rows most relevant to each sub-question.
- A language model writes the answer. The model combines those passages into a response.
- Citations are attached. Each numbered footnote points to a page that supplied part of the answer.
The practical lesson: you are competing at the passage level, not the page level. A 3,000-word article with one buried, vague answer can lose to a 600-word page with one crisp paragraph that answers the exact sub-question.
Which signals seem to decide who gets cited?
Relevance to the specific sub-question comes first, then whether the page looks trustworthy and current enough to quote. No one outside Perplexity knows the weights, but these factors line up consistently with which pages appear in its footnotes.
| Signal | What it looks like on a page | How much you control it |
|---|---|---|
| Direct relevance | A heading and paragraph that match the question almost word for word | High |
| Extractability | Short answer paragraphs, lists, tables, clear labels | High |
| Crawl access | PerplexityBot allowed, content in the HTML, no login wall | High |
| Search visibility | The page already ranks reasonably in traditional search | Medium |
| Freshness | Visible dates, current prices and versions, recent updates | High |
| Source reputation | The site is known and referenced for this topic | Medium, slow to build |
| Specificity | Numbers, named processes, concrete examples rather than generalities | High |
Notice how many rows you control directly. Reputation takes time, but extractability, freshness and access are things you can fix this week.
Can Perplexity even see your website?
Many sites never get cited simply because Perplexity can't read them properly. Check this before you touch any content.
- robots.txt. Perplexity documents a crawler called PerplexityBot. If your robots.txt blocks it, or blocks all unknown bots by default, your pages are unlikely to make it into its index. Some security plugins and CDN bot-protection settings block AI crawlers without you knowing.
- JavaScript-only content. If your text only appears after heavy client-side JavaScript runs, some crawlers see a nearly empty page. Content should be in the HTML the server sends.
- Gated pages. Pricing behind a form, articles behind a login or answers inside PDFs that aren't linked anywhere are effectively invisible.
- Slow or failing pages. A crawler that times out moves on. On grww.tech we cut portfolio image weight from 1.66 MB to 456 KB by switching to WebP and load analytics after the page renders, partly so every bot and visitor gets the content fast.
If you only do one thing after reading this, open your robots.txt and confirm you aren't blocking the AI crawlers you want citations from. Blocking is a legitimate choice for some publishers, but it should be a decision, not an accident.
What kind of page does Perplexity like to quote?
Pages that answer one question per section, in the first sentence, with something specific. Perplexity's answers are short, so it favours passages that already read like a good answer.
Compare two ways of answering "how long does it take to build a small business website?":
- Weak: "Timelines vary depending on many factors, and every project is unique, so it's best to talk to an expert."
- Strong: "A marketing website typically takes 1–2 weeks once content is ready; e-commerce or web apps take up to about 4 weeks."
The second version is quotable on its own. It names a scope, gives a range and states the dependency. That pattern is drawn from how we phrase our own published FAQ, and it is the kind of sentence that survives being cut out of its page.
Page types that tend to earn citations:
- Explainers with question-style headings that mirror how people phrase follow-ups.
- Comparison pages with a table that summarises the trade-offs.
- Pricing and cost guides that give honest ranges with clear assumptions.
- Original data, even small datasets from your own work, because nobody else can supply it.
- Definitions and glossaries for terms in your industry.
How is Perplexity different from Google AI Overviews and ChatGPT?
All three blend search with a language model, but Perplexity leans hardest on live retrieval and shows its sources most prominently. That changes what you optimise for.
| Perplexity | Google AI Overviews / AI Mode | ChatGPT (with search) | |
|---|---|---|---|
| Main source of facts | Live web retrieval for nearly every answer | Google's own index | Mix of training data and web search |
| Citation visibility | Numbered footnotes on almost every claim | Linked cards and inline links | Inline links when search is used |
| Crawler to allow | PerplexityBot | Googlebot | OAI-SearchBot for search features |
| What helps most | Clear passages, freshness, access | Strong classic SEO plus clear answers | Brand presence across the web plus search visibility |
For the platform-specific details, see our guides on optimising for Google AI Mode and getting your brand recommended by ChatGPT. The overlap is large: a page that is easy to extract and trustworthy tends to do well in all three.
A 30-day plan to become a Perplexity source
Start with access, then rewrite your most commercially important pages for extraction, then add things only you can say. Here is a realistic sequence for a small team.
Week 1: access and baseline.
- Check robots.txt, CDN bot rules and security plugins for PerplexityBot blocks.
- Make sure key pages render their text without JavaScript.
- Ask Perplexity 15–20 questions your customers actually ask and note which sites it cites. Those are your real competitors for answers.
Week 2: restructure five pages.
- Pick the five pages that matter most to revenue: core service pages, a pricing guide, a comparison.
- Rewrite each H2 as a question and open each section with a one- or two-sentence answer.
- Add a summary table where there is a comparison or a set of options.
Week 3: add specificity.
- Replace vague claims with ranges, timelines, process steps and examples.
- Add visible "last updated" dates and refresh outdated figures.
- Include a short, factual "about us" block with who you are, where you work and what you've delivered.
Week 4: build reputation signals.
- Publish one piece of original data from your own work, anonymised.
- Make sure your business details match across your site, Google Business Profile and directory listings.
- Re-run the same question set and record changes.
The beginner's guide to answer engine optimisation covers the writing side in more depth and explains why classic rankings still feed into all of this.
When is chasing Perplexity citations not worth it?
If your customers don't use Perplexity, or your business wins work almost entirely through referrals and local search, it shouldn't be your first priority. A local plumber will usually get more from a well-run Google Business Profile than from Perplexity citations.
It is worth more for businesses whose buyers research carefully before contacting anyone: software buyers, B2B services, consultancies, education and immigration advisers. The good news is that the work overlaps heavily with what makes pages rank and convert anyway, so effort spent here rarely goes to waste.
Frequently asked questions
Does Perplexity pay attention to Google rankings?
Perplexity runs its own retrieval and doesn't publish whether it uses Google results directly. In practice, pages that rank well in traditional search are often the same pages it cites, because the qualities that earn rankings, relevance, authority and clarity, overlap with what makes a passage worth quoting.
Should I block PerplexityBot to protect my content?
That depends on your business model. Publishers who sell content sometimes block AI crawlers deliberately. For most service businesses, being cited is free exposure to people researching a purchase, so blocking usually costs more than it protects.
How long does it take to start appearing as a source?
There is no fixed timeline. Because Perplexity searches live, changes to an already-indexed page can show up relatively quickly, while a brand-new site with no search visibility or reputation will take longer. Track a fixed set of questions monthly rather than checking daily.
Do I need special markup or files for Perplexity?
No special file or tag is required. Clean HTML, sensible headings and accurate structured data help any machine reader understand your page, but the deciding factor is still whether your content answers the question clearly.
Can I pay to be cited by Perplexity?
Perplexity has experimented with advertising formats, but organic citations within answers are not something you can buy. Any agency claiming it can guarantee citations is overpromising.
Want your pages to be the ones AI answers quote?
We restructure service pages, pricing guides and comparison content so both search engines and answer engines can understand and cite them, drawing on 170+ websites shipped across four markets. If you'd like us to audit your top pages and show you exactly what to change, ask for a fixed-price quote.