Guides / Perplexity

How to get your SaaS recommended by Perplexity

By Highpoint Labs, LLC · Published October 27, 2026 · 6 min read

Is AI recommending you or your competitors?
Free check: real ChatGPT answers to buyer questions in your category. No account needed.
Run a free check

Perplexity is built around search. Ask it "what's the best scheduling software for consultants?" and it searches the web, writes an answer, and shows numbered citations next to the claims. For a SaaS team that's good news: the sources are visible, so you can see exactly which pages shape the answer and work on them.

This guide covers how Perplexity finds and cites pages, the practical steps that improve your odds of being recommended, and how to measure progress. A note on our data: ShortlistWatch's published AI rankings and free check currently measure ChatGPT, not Perplexity. Where we quote numbers below, they come from our ChatGPT runs and are there to illustrate how search-based answers behave, not as Perplexity measurements.

How Perplexity picks sources

Perplexity's documentation describes two automated visitors:

  • PerplexityBot crawls and indexes the web "to surface and link websites in search results on Perplexity". Perplexity says it is not used to crawl content for AI foundation models, and recommends allowing it in robots.txt (and allowing its published IP ranges) if you want to appear.
  • Perplexity-User visits pages when a user's question needs them, and Perplexity says it generally ignores robots.txt because a person requested the fetch.

In practice, Perplexity answers a question by finding pages relevant to it, reading them, and summarizing what they say, with citations. That has three consequences for a SaaS team:

  1. If Perplexity can't crawl your pages, it's much less likely to cite them.
  2. The answer is shaped by whichever pages it retrieves, which for "best X" questions are usually third-party roundups, comparison pages, directories, review sites, discussion threads and vendor pages.
  3. Because citations are shown, you can audit the answer. Nothing has to be guessed.

Step 1: Let PerplexityBot in

Check three places:

  1. robots.txt. Make sure PerplexityBot isn't disallowed, either by name or by a blanket rule for all user agents.
  2. Your CDN or WAF. Perplexity's docs include instructions for allowlisting its bots in Cloudflare and AWS WAF, because a firewall can block it even when robots.txt allows it. If you use bot-fight or challenge modes, check your logs for PerplexityBot requests being blocked.
  3. Rendering. Make sure your key pages (homepage, pricing, features, comparisons, docs) serve their main text in the HTML.

Perplexity notes that robots.txt changes can take up to 24 hours to take effect.

Step 2: Run the audit only Perplexity makes easy

Because every Perplexity answer lists its sources, the fastest research you can do takes an hour:

  1. Write 10 to 20 questions a buyer in your category would ask: "best [category] for [audience]", "[competitor] alternatives", "cheapest [category] with [feature]", "[you] vs [competitor]".
  2. Ask each one in Perplexity and record which products are named, in what order, and every cited URL.
  3. Group the cited URLs by type: editorial roundups, review platforms, directories, forums, vendor sites, documentation, video.
  4. Mark which ones mention you, which mention competitors but not you, and which mention nobody specific.

The middle group, pages that cite competitors but not you, is your work list.

In our ChatGPT data the cited sources were spread thin: 100 answers across 20 categories linked 424 different websites, 204 of them cited inline to back a specific claim. Expect your category's list to be long and specific to it, and work through the sites that recur.

Step 3: Earn a place on the pages it cites

For each recurring source type, there's a legitimate way in:

  • Editorial "best of" lists. Find the author and pitch a short, specific note: what you do, who it's for, how you differ, and a free account to test. Offer to correct outdated facts about competitors only if you're sure; writers appreciate accuracy, not lobbying.
  • Review platforms such as G2, Capterra or TrustRadius. Claim your profile, fill in categories and pricing, and ask real customers for reviews at natural moments (after onboarding, after a support win). Don't incentivize only positive reviews; it breaks the platforms' rules.
  • Directories and marketplaces. Complete profiles on the directories your category actually uses, plus integration marketplaces (Zapier, Slack, Shopify and so on) if you integrate with them. These pages often describe your product in plain category language.
  • Community discussions. If threads on Reddit, Hacker News, Indie Hackers or niche forums show up in your audit, take part as yourself: answer questions, disclose that you work on the product, and don't astroturf. Communities punish fake accounts, and so will your reputation.
  • Your own documentation. Help docs and integration guides answer the specific "can X do Y?" questions buyers ask. Keep them public and crawlable.

Step 4: Make your facts consistent

Perplexity summarizes whatever it retrieved. If three sites describe your pricing three different ways, the answer may pick the wrong one, or hedge and skip you.

  • Use one product name and one category label everywhere. If you renamed, put "formerly [old name]" on your homepage and update old profiles. In our ChatGPT data, an answer that cited no sources still recommended "Sendinblue" and "ConvertKit", names those companies have since replaced. Stale names linger.
  • Keep a single, current source of truth for pricing, plan limits and integrations on your own site, and update directory profiles when it changes.
  • Write a plain positioning sentence ("[Product] is [category] for [audience] that [difference]") and reuse it in profiles and bios.
  • Add structured data (Organization and SoftwareApplication with sameAs links to official profiles) that matches what's visible on the page.

Step 5: Write pages that answer the question asked

Perplexity cites pages that answer the question. For buyer questions, that means:

  • Comparison pages ("[You] vs [Competitor]"). Be fair, include a quick summary table, and say who should pick the competitor. Balanced pages are more useful and more quotable.
  • Alternatives pages ("[Competitor] alternatives"). Include several real options, not just you.
  • "Best for" pages for the segments you really serve, such as agencies, freelancers or startups under 10 people.
  • Clear headings and direct answers. Put the answer in the first sentence under each heading, then the detail. It helps readers skim and makes the relevant passage easy to quote.
  • Dates and specifics. Show when a page was last updated and use concrete facts (plan prices, limits, supported integrations) instead of adjectives.

Specific questions favor specific pages. In our ChatGPT email marketing data, a question asking for an "affordable email marketing software with good support" surfaced five smaller products that none of the other four answers in that category mentioned. Smaller products tend to win on narrow questions, so write for them.

Step 6: Track it over time

Single answers are noisy. Our ChatGPT runs show how noisy: of 264 product-and-category pairs we recorded on 26 October 2026, 151 (57%) appeared in only one of the category's five answers, and only 9 products appeared in all five. Expect Perplexity answers to vary too, by wording, time and model.

A simple tracking routine:

  1. Keep a fixed list of buyer questions and re-ask them on a schedule (weekly or every two weeks).
  2. Record, per answer, whether you're named, your position, and the cited sources.
  3. Compare your share of answers with your two or three closest competitors, not with the category leader.
  4. Log your actions (a listicle placement, a new comparison page, a directory profile) so you can see whether the cited sources change afterwards.
  5. Check your referral analytics for perplexity.ai traffic. It's a direct signal that citations are turning into visits.

Common mistakes

  • Blocking AI crawlers by accident with a security setting, then wondering why you're never cited.
  • Treating one screenshot as the truth. Ask several times, in several phrasings.
  • Optimizing only your homepage. Most citations for "best X" questions point to third-party pages.
  • Gaming communities or reviews. Short-term mentions, long-term damage.

The short version

  1. Allow PerplexityBot in robots.txt and in your firewall.
  2. Ask your buyers' questions in Perplexity and list every cited source.
  3. Get honestly included on the recurring sources that mention competitors but not you.
  4. Keep your name, category, pricing and positioning identical everywhere.
  5. Publish fair comparison, alternatives and "best for" pages with direct answers.
  6. Re-run the same questions on a schedule and track share, position and sources.

Most of this also helps with ChatGPT and Google Gemini. If you want to see where you stand in ChatGPT today, the free check runs real buyer questions for your category and shows who gets named.

See where you stand today

The free check asks ChatGPT the questions your buyers ask, for your product and the competitors you name, and shows who gets named, in what position, and which sources are cited. Then track it weekly.

Run a free check