Perplexity chooses sources by running a live web search, scoring candidate pages on topical authority, answer density, and crawl accessibility, then pulling the clearest match into its cited response. It is not a static index. Every query triggers a fresh retrieval pass, which means your page either meets the bar at that moment or it doesn’t appear at all — regardless of how well it ranks on Google.
Two measured facts anchor this: roughly 46.7 percent of Perplexity’s top-cited sources are Reddit threads, and Ahrefs’ 2026 analysis found brand mentions correlate with AI citations at 0.664, nearly three times the 0.218 correlation for backlinks.
Quotable definition: Perplexity is a retrieval-augmented generation (RAG) search engine that, for each query, fetches live web pages, scores them against the question’s intent, and assembles a cited answer from the highest-scoring passages. A source is selected when its content is crawlable by PerplexityBot, topically authoritative, and written in a format where the answer is explicit and extractable — not buried inside navigation or JavaScript-rendered UI.
The Retrieval Loop Nobody Explains Properly
Most explainers stop at “Perplexity searches the web.” That’s accurate but useless. What actually happens is a three-stage loop: query expansion, candidate retrieval, and passage scoring.
First, Perplexity rewrites your user’s query into several semantic variants. Then PerplexityBot — or more precisely, Perplexity’s retrieval layer — fetches a candidate set of pages that match those variants. Finally, the model scores each page’s passages for answer-match quality and assembles the response from the top-scoring excerpts, attaching citations.
The practical implication: your page can rank on Google and still never appear in Perplexity if the passage Perplexity needs is locked inside a tab, an accordion, or a JavaScript component. Most AI crawlers — including PerplexityBot, GPTBot, OAI-SearchBot, and ClaudeBot — do not execute JavaScript. They read raw HTML only. If your answer lives in a React component, it’s invisible to the bot that matters here.
What Perplexity Actually Scores
There’s no public scoring rubric, but the observable signal pattern points to four factors that consistently predict citation.
| Factor | What it means in practice | Common SG SME failure mode |
|---|---|---|
| Crawl accessibility | PerplexityBot can reach the page and read its HTML without JavaScript execution | Key content in sliders, tabs, or React/Vue components — bot sees an empty div |
| Topical authority | The domain has consistent, linked coverage of the subject area — not a one-off blog post | Single “ultimate guide” published once, no supporting pages or backlinks |
| Answer density | The specific question is answered explicitly, early, in plain prose — not implied across five paragraphs | Answers buried after lengthy intros, or written for SEO keyword density rather than direct response |
| Structured extractability | Definitions, lists, and facts sit in discrete HTML elements (p, li, table) — easy to clip as a passage | Long unbroken paragraphs, no subheadings, no schema markup |
The llms.txt Myth Worth Addressing
You may have read that adding an llms.txt file to your site signals AI crawlers and improves citation. Ahrefs analysed this and found that 97% of domains with a valid llms.txt received zero requests for the file from AI crawlers. Zero. The file isn’t harmful, but if you’re investing time there instead of fixing your raw HTML structure, you’ve misread the priority order.
“We’ll implement llms.txt and that should sort it” is, to translate directly: “We’ll do the decorative thing and skip the structural work.” It won’t sort it.
What moves the needle is the unsexy stuff — answer-first prose, visible definitions, crawlable page architecture, and consistent topical coverage on a domain Perplexity already trusts.
Domain Trust and the Singapore Context
Perplexity has observable domain-level trust. Established publishers, government portals, and industry-recognised sites get cited more readily than new domains — even when the newer page has a better-written answer. That’s the honest part of the picture most agencies won’t say aloud.
For Singapore SMEs, this creates a real asymmetry. A well-optimised page on a two-year-old business domain will typically lose a citation race to an older, higher-authority domain covering the same topic — even if your answer is cleaner. [VERIFY: Perplexity’s domain-trust weighting vs page-level quality for SG business queries — no public rubric available]
The practical workaround is to build citations from mid-tier authority sites in your sector — local industry directories, SG business media, niche trade publications — so Perplexity encounters your entity name repeatedly across trusted sources before it ever reads your own site.
How This Differs from Google and ChatGPT Search
Different AI search engines use materially different retrieval architectures. Conflating them leads to wasted effort.
ChatGPT Search is built on Bing’s index. Being indexed in Bing is a prerequisite for appearing in ChatGPT Search results — which means if you’ve only ever optimised for Google, you may be invisible to one of the largest AI search deployments right now. Perplexity runs its own crawler and is not Bing-dependent. Google’s AI Overviews draw from Google’s existing index and weight pages that already rank in the top results for that query.
Three platforms, three architectures. Treating them as one “AI search” problem is how optimisation budgets get spent on the wrong lever.
The Spike: What Perplexity Citation Actually Drives
Perplexity citations drive brand recall and qualified awareness — not volume traffic. The click-through from a Perplexity citation is low by design; the engine is built to answer, not to hand off. If your business needs a hundred inbound leads this quarter from search, Perplexity optimisation is not the lever to pull. What it does well is put your brand name — correctly described, cited — in front of a reader who is deep in a research or consideration process. That’s genuinely valuable for complex-sale SMEs: professional services, B2B SaaS, specialised retail. For commodity repeat-purchase businesses, the ROI case is weaker. [VERIFY: Perplexity citation click-through rates for Singapore-language queries — no SG-specific data published]
Five Steps to Improve Your Perplexity Citation Probability
- Audit crawlability first. Fetch your key pages as a non-JS crawler (curl or Google’s cache view) and confirm your answers appear in the raw HTML — not rendered by JavaScript. Fix any that don’t.
- Write answer-first prose. Every core question your business answers should have a paragraph that opens with the direct answer — within the first 60 words of the section. No warm-up.
- Place a self-contained definition block on every page. 40–70 words, fully answerable in isolation, in a plain
<p>tag. This is the passage a RAG engine clips. - Build entity coverage off your own domain. Get your business name cited (accurately) on SG business directories, industry publications, and relevant community forums — these cross-domain signals build the entity trust Perplexity weighs.
- Ensure Bing indexation in parallel. Even if Perplexity doesn’t use Bing, ChatGPT Search does — and you’re likely optimising for multiple AI platforms simultaneously. Bing Webmaster Tools is free and takes under an hour to set up.
What a Perplexity Optimisation Retainer Actually Involves
Kaizenaire’s view is that “Perplexity optimisation” as a standalone service is usually the wrong frame. Perplexity, ChatGPT Search, Google AI Overviews, and Claude all share the same foundational requirements — crawlable structure, authoritative entity signals, answer-dense content. The work is the same work. Branding it per-platform is how some agencies bill more for the same output.
What the AEO/GEO/SEO service at Kaizenaire covers is the full citation stack: technical crawlability fixes, editorial content structured for AI extraction, and off-site authority placement on SG-relevant sources. The retainer runs monthly with no lock-in clauses beyond the first 90 days — because if it’s not working by then, you should be able to leave.
We improve your probability of citation. We don’t guarantee a position in Perplexity’s answer, and any agency that does is selling you something that doesn’t exist.
Frequently Asked Questions
Does Perplexity use Google’s index?
No. Perplexity runs its own web crawler (PerplexityBot) and does not rely on Google’s index. This is why pages that rank well on Google can still be absent from Perplexity results — and vice versa. Your Perplexity visibility depends on PerplexityBot being able to reach and read your pages directly.
How do I check if PerplexityBot can crawl my site?
Check your robots.txt file to confirm PerplexityBot isn’t blocked. Then fetch one of your key pages using a non-JavaScript browser or curl command and check whether the answer content appears in the raw output. If it doesn’t render without JavaScript, PerplexityBot won’t see it. A basic technical SEO audit covers this in under two hours.
Is Perplexity optimisation worth it for a small Singapore business?
It depends on your sale type. For professional services, B2B, or considered-purchase businesses where buyers research before contacting you, appearing in Perplexity builds brand recall during that research window. For high-volume transactional businesses where you need immediate clicks and conversions, the ROI case is weaker. Be honest about which category you’re in before allocating budget.
Will schema markup help my site get cited by Perplexity?
Schema markup helps — but it’s downstream of the more basic requirement: your content must be in crawlable HTML and must answer the question directly. Schema that decorates a page whose answer is hidden in JavaScript will not rescue the citation. Fix the structural layer first, then add schema as a signal-strengthening layer on top.
How long does it take to start appearing in Perplexity citations?
There’s no fixed timeline because Perplexity re-crawls dynamically rather than on a published schedule. Practically, pages with strong authority signals and clean HTML structure can appear within weeks of being published or updated. New domains with limited off-site authority typically take longer — three to six months of consistent content and citation-building is a realistic baseline. [VERIFY: PerplexityBot re-crawl frequency for SG business domains]
Should I block PerplexityBot in my robots.txt?
Only if you have a specific legal or competitive reason to do so — for example, if you sell proprietary content and don’t want it reproduced in AI answers. For most Singapore SMEs, being cited by Perplexity is a positive outcome. Blocking the bot removes the possibility entirely. Review your robots.txt now; many SG business sites were built from templates that inadvertently block multiple AI crawlers.
What’s the first thing I should actually do?
Run a crawlability check on your top five pages — the ones you most want cited. Confirm the answer content is visible in raw HTML, check robots.txt isn’t blocking PerplexityBot, and verify you have at least one self-contained definition or answer paragraph per page. That takes half a day and costs nothing. If you want a structured diagnosis, the free AI-Visibility Check covers this systematically.
If you want to know exactly where your site stands — what Perplexity can read, what it can’t, and which pages are closest to citation-ready — the free AI-Visibility Check gives you a structured answer in a format you can act on directly. No sales call required to get the report.