How to rank in Perplexity — Sonar retrieval explained (2026)

Perplexity is a retrieval-first AI search engine. Every answer is grounded in live web fetches — there is no significant "answer from training" mode like ChatGPT has. The retrieval pipeline (called "Sonar") works roughly like this:

1. Query expansion. Perplexity rewrites your prompt into ~3–5 sub-queries.
2. Retrieval. Each sub-query hits the open web. Sonar fetches approximately 10 candidate URLs per query (per the Growth Marshal Sonar Playbook 2025).
3. Re-ranking. The retrieved set is ranked by source authority, freshness, snippet quality, and semantic relevance to the sub-query.
4. Citation selection. The top 3–4 URLs become the visible citations. The answer prose synthesises facts from those URLs and includes inline footnote links.

Practical implication: the funnel is retrieve 10 → cite 4. You don't need to be the #1 source — you need to be in the top 4 for the retrieval step to pick you, which is a materially different game from classic Google ranking.

The single most cited domain in Perplexity in 2026 is Reddit, at 24% of all citations (Tinuiti Q1 2026 AI Citation Trends Report). Three reasons:

• Sonar weights authority + freshness. Reddit threads accumulate authority via upvotes and cross-linking, and a new thread on the same topic can outrank an old static page within days.
• Reddit text is high-density.A 200-word Reddit comment debating "X vs Y" contains more factual claims per token than most product pages. The retrieval model picks it preferentially.
• Perplexity licenses Reddit data. Per OpenAI and Google's reported Reddit deals (and Perplexity's own confirmations), Reddit content is available to Perplexity via API, not just crawl. Reach and recency are higher than for sites it has to crawl.

Practical implication: if your brand isn't mentioned in any of the top-upvoted Reddit threads about your category, you are missing the single highest-leverage citation lever Perplexity offers. See Reddit citation strategy for the tactical playbook.

• PerplexityBot— the main crawler. Builds and refreshes Perplexity's search index. Block to be excluded entirely.
• Perplexity-User— live retrieval triggered by a user query. Per Perplexity's documentation, user-initiated fetches sometimes ignore robots.txt because the user explicitly requested the fetch.

If you want to be cited by Perplexity, allow both. If you want to opt out, block both — and also rate-limit at the CDN level for Perplexity-User since polite-bot rules may not apply.

Five signals Sonar weights heavily (per public Perplexity statements + observed behaviour):

• Source authority. Domain Rating / authority equivalents matter. A new domain with thin backlinks rarely makes the top-4 cut.
• Freshness. Content updated within the last 30 days is materially more likely to be picked over a 12-month-old page on the same topic. Date metadata visible to crawlers matters.
• Snippet density. A page with one high-value paragraph that directly answers the query beats a 3000-word marketing essay where the answer is buried in paragraph 14.
• Schema.org markup. Sonar extracts structured facts (price, availability, ratings, FAQ answers) preferentially from JSON-LD blocks vs inferring them from prose.
• Citation footprint elsewhere. Pages that are themselves cited / linked by Reddit, YouTube, Wikipedia, and high-authority blogs get a boost when Sonar retrieves them — like classic PageRank but weighted toward AI-friendly source types.

1. Allow both PerplexityBot and Perplexity-User in robots.txt. Use our robots.txt builder if you want to opt out of training while staying citable.
2. Rewrite your top 5 pages for snippet density. Move the answer to the first 200 words. Use clear H2s matching common buyer questions. One claim per sentence beats one sentence per paragraph.
3. Ship Organization, SoftwareApplication, and FAQPage JSON-LD. Sonar extracts these preferentially. Our JSON-LD generator covers all three.
4. Set up Reddit presence in your category. Find the 3–5 subreddits where your buyer-question gets asked. Comment with disclosure on threads — value first, no link spam. This is the highest-leverage move.
5. Publish a vendor-neutral comparison page. "X vs Y vs Z" pages with honest pros/cons get cited disproportionately by Sonar when users ask category comparison questions.
6. Refresh the same 5 pages every quarter. Sonar prefers content updated within 30 days. Even a small update + new dateModified bumps your retrieval probability.
7. Get listed on a third-party comparison site. G2, Capterra, alternative-to.net — these get retrieved often. A profile with reviews counts as a citation pointer to your brand.
8. Measure Share of Voice on a buyer-question panel. Free 60-second audit at /grader runs this exact panel through Perplexity Sonar.

• Treating Perplexity like Google.Sonar retrieves 10 and cites 4. Being "ranked 8" in Google means nothing — you either make the top 4 of Sonar's retrieval set or you don't.
• Ignoring Reddit.24% of citations. Single highest-leverage source. Most B2B SaaS founders treat Reddit as "not serious" and skip it — this is a Perplexity gift.
• Buried answers.If the answer to "what does this product cost" is in paragraph 9, the snippet won't make it into the retrieval re-ranking. First-200-words rule applies.
• No freshness signal. Static pages from 2024 lose to fresh pages from last month. Update or lose citation probability.

1. Ask Perplexity directly on your fixed buyer-question panel. Note which 3–4 sources got cited. Re-run weekly — the citation set drifts.
2. Run our free audit. /grader runs a buyer-question panel through Perplexity Sonar and returns the cited brands plus your Share of Voice in 60 seconds.
3. Track Share of Voice over 30 days. Single scans are noise — trend over weeks is signal.

1. Reddit citation strategy — the tactical how-to for the highest-leverage Perplexity source.
2. How to get cited by ChatGPT — different retrieval, more conservative authority bias.
3. How to measure AI visibility — close the loop on what you shipped.