Perplexity SEO is the work of getting your pages cited, and your brand named and recommended, in Perplexity's answers. Perplexity searches the web in real time using its own crawler and index, and it puts numbered citations on every answer. To become one of them, let PerplexityBot crawl your site, publish pages that answer your buyers' questions directly with specifics, get your brand onto the sources Perplexity already cites for your topic, and measure across many runs, because the answer changes.
Nobody can guarantee a citation. Perplexity publishes no ranking factors, and its one instruction to site owners, in its crawler documentation, is to allow PerplexityBot and its published IP addresses.
I reread Perplexity's documentation in October 2026. The Perplexity page at the top of Google for "perplexity seo," Improve SEO strategy, is a product page about using Perplexity to research topics and competitors for a content calendar. It says nothing about how Perplexity picks sources.
How does Perplexity choose its sources?
Here's what Perplexity's help center, crawler docs and the 2025 launch post for its Search API, which Perplexity says shares the answer engine's infrastructure, do say.
- It searches for each question, "in real-time, gathering insights from top-tier sources," then writes the answer with a model such as GPT-5 or Claude 4.6 Sonnet.
- It searches its own index, which Perplexity says covers "hundreds of billions of webpages." PerplexityBot feeds it, and Perplexity also "partners with third-party crawlers to help build our search index," according to its robots.txt help article.
- It scores parts of pages. Its infrastructure "divides documents up into fine-grained units" that are "individually surfaced and scored against the original query."
- It may fetch a page live. Perplexity-User "might visit a web page to help provide an accurate answer and include a link to the page in its response."
- It cites. "Each answer includes numbered citations linking to the original sources." Some carry a Government, Academic or Trusted source label, but Perplexity says "Most domains on the web do not have a label" and doesn't say labels affect which sources it picks.
What Perplexity doesn't say: how it ranks pages, whether Google or Bing results feed it, or why it cites one page over another. Claims about Perplexity ranking factors come from outside studies that observe answers, which show association, not cause.
PerplexityBot and Perplexity-User: what blocking changes
| Agent | What it does, per Perplexity | Follows robots.txt? | If you block it |
|---|---|---|---|
| PerplexityBot | "designed to surface and link websites in search results on Perplexity." Not used to train models. | Yes | Your page text isn't indexed, but Perplexity "may still index the domain, headline, and a brief factual summary." |
| Perplexity-User | Visits a page when a user's question calls for it, and may link to it. | "generally ignores robots.txt rules" | A robots.txt rule probably won't stop it. Perplexity doesn't say what a firewall block changes. |
Perplexity's pages disagree on Perplexity-User: the help center says a feature that summarized robots-blocked URLs "has been disabled to prevent misuse," while the developer docs still say the fetcher ignores robots.txt. Perplexity says it "does not build foundation models," so there's no training crawler to weigh against search. Allow both agents.
The Cloudflare dispute
In August 2025, Cloudflare said that after it blocked Perplexity's declared agents on test domains, Perplexity still answered questions about them, using an undeclared agent posing as Chrome from IP addresses outside Perplexity's published ranges. Cloudflare removed Perplexity from its verified bots list. Perplexity replied that Cloudflare had confused it with traffic from BrowserBase, a cloud browser service it "only occasionally uses," and that fetching a page for a user isn't crawling.
Neither side published data anyone can check. Cloudflare mentioned the de-listing again in October 2026, and I couldn't confirm Perplexity's current status, so don't assume Cloudflare lets PerplexityBot through.
How Perplexity differs from ChatGPT and Google
| For a site owner | Perplexity | ChatGPT search | Google AI Overviews and AI Mode |
|---|---|---|---|
| When it searches | For each question | When a question "would benefit from current information," or when the user turns on Search, per OpenAI | Inside Google Search |
| Sources on screen | Numbered citations on every answer | Answers that use search "may include citations" | Supporting links beside the AI answer |
| Where pages come from | Its own index, built by PerplexityBot and partner crawlers | Third-party search providers, with Microsoft named, plus OAI-SearchBot | Google's Search index |
| If you block its crawler | Domain, headline and a brief summary may stay | Navigational links may stay, per OpenAI | No AI-only robots.txt token. A Search Console setting can exclude a site. |
| Report for site owners | None I could find | None I could find | Search Console's generative AI performance report |
Three differences change the work. The citations are always on screen, so the buyer sees which pages Perplexity relied on, and so do you. The mix leans more on social platforms: a University of Toronto preprint found 67.4% of the links Perplexity cited for well-known brands were earned media, meaning other publishers' pages, and 23.8% were social platforms, against 93.5% and 0% for ChatGPT.
And Perplexity doesn't say Google or Bing rankings feed its index, so check PerplexityBot's access on its own. Its source mix was also steadier than ChatGPT's in 2025, according to Semrush.
How to get cited by Perplexity, step by step
People search for how to rank on Perplexity, but there's no fixed rank to win. What you can raise is the share of answers that name you, recommend you and cite your pages.
Step 1: Let PerplexityBot in
Perplexity's crawler docs recommend "allowing PerplexityBot in your site's robots.txt file and permitting requests from our published IP ranges." Open yourdomain.com/robots.txt. If it blocks PerplexityBot by name, or every crawler with User-agent: *, add:
User-agent: PerplexityBot
User-agent: Perplexity-User
Allow: /A crawler with its own group ignores the * group, so copy over any paths you want kept private. Then check what sits in front of your site:
- In your firewall, allow requests whose user agent contains PerplexityBot or Perplexity-User and whose IP is on Perplexity's lists for PerplexityBot and Perplexity-User, as its docs recommend. The lists change, so update the rule automatically.
- On Cloudflare, set Search to Allow in AI bot policies, make sure the Agent setting, which covers "chat fetch bots," isn't blocking Perplexity-User, and confirm PerplexityBot is allowed in AI Crawl Control.
- In your server logs, look for PerplexityBot requests that get a 200, not a 403 or a challenge page.
Perplexity doesn't say whether PerplexityBot runs JavaScript, so put prices and product details in the HTML your server sends. Ready-made files for every AI crawler are in AI crawlers and robots.txt.
Step 2: Pick the questions to check
List 10 to 30 questions buyers ask before choosing a product like yours: "best X for Y," comparisons, alternatives, pricing and how-to questions. Take the wording from sales calls, support tickets and Search Console, and keep the context buyers add, such as team size or budget. The guide to getting cited by ChatGPT shows where to find each type.
Step 3: Check Perplexity by hand, more than once
- Ask each question in a new thread, since Perplexity remembers context within a conversation.
- Use the default search with the model on "Best." If you test Pro Search or Research, record them separately.
- Run each question five times over a few days.
- For each run, record whether you were named, recommended or cited to your own page, plus the competitors and every numbered source.
Why five runs? Envoyra, which sells AI visibility tracking, asked four assistants, including Perplexity with web search on, 170 questions a week for four weeks, and 79% of question and assistant pairs changed their brand list at least once.
Here's the sheet, as an invented example with placeholder names, for "best project management tool for a 15-person agency":
| Run | Named | Recommended | Your page cited | Competitors | Sources cited |
|---|---|---|---|---|---|
| 1 | Yes | No | No | A, B | Reddit thread, review site |
| 2 | Yes | Yes | Yes, pricing page | A | Your pricing page, Reddit thread |
| 3 | No | No | No | A, C | "Best tools" list, YouTube review |
| 4 | Yes | No | No | A, B | Reddit thread, A's comparison page |
| 5 | No | No | No | A, B | Reddit thread, review site |
Your Brand was named in 3 of 5 runs, recommended in 1 and cited in 1. The same Reddit thread appeared in 4 runs, so it's the first outside source to work on.
If you'd rather not do this by hand, you can check what Perplexity says about you with Inkie's free report on 5 buyer questions. I make Inkie; see the tools section below.
Step 4: Build the page that answers each question
For each question, pick the one page that should answer it. If none exists, that's the gap. If one exists, compare it with the pages Perplexity cited instead: what do they say that yours doesn't?
Put the direct answer near the top, then the specifics: the price at the buyer's size, who the product suits and who it doesn't, and a "prices as of" date. "Affordable plans for every team" gives Perplexity nothing to quote. The per-user price, the user limit and what costs extra do.
Make each section answer its own heading, and add bylines. Perplexity scores "fine-grained units" of pages, so by my reading a plainly answered section should match more easily, and its source review asks whether a site says who wrote each piece. Neither is Perplexity's advice for getting cited. Don't chop pages into fragments, which Google's Danny Sullivan warns against.
Keep pages current. Ahrefs, which sells SEO tools, found pages cited by AI assistants, Perplexity among them, averaged 1,064 days old against 1,432 for organic results. Update when facts change, but never just the date: in a peer-reviewed study, a false "published on" date moved passages up as many as 95 places when language models reranked them.
Step 5: Get onto the sources Perplexity already cites
Your sources column shows where Perplexity gets its picture of your category. Reddit stands out, though vendors disagree on how much: Profound counted it as 6.6% of Perplexity citations, and Semrush found Reddit links in 3.5% of Perplexity answers. They measured different things in different months, so trust your own sheet.
For commercial questions, list articles matter. A Wix and Peec AI study of citations from ChatGPT, AI Mode and Perplexity found list articles made up 40.86% of citations for commercial prompts. Both companies sell to marketers.
| Source | What to do | What not to do |
|---|---|---|
| Reddit threads | Answer under your real name, say where you work, follow each subreddit's rules. Semrush found most cited threads had fewer than 20 upvotes. | Use alternate accounts or fake customers. |
| YouTube | Publish demos and honest comparisons. | Upload a pitch that answers nothing. |
| Review sites such as G2 | Keep your profile's pricing current. Ask real customers for reviews. | Buy, swap or write reviews. |
| "Best X" lists | Send authors accurate facts and ask for corrections. | Pay for undisclosed placement or rank yourself first. |
A citation still isn't a recommendation. In an Ahrefs experiment across four assistants including Perplexity, 43% of answers citing Ahrefs' pages about its own conference never mentioned the conference.
Step 6: Measure runs and referral visits
Re-run the same questions weekly with the same settings, and track the share of runs that name you, recommend you and cite your page, and your main competitors' shares. A 2026 preprint that sampled Perplexity repeatedly found "many apparent differences between domains fall within the noise floor," so small samples can't show small changes. Perplexity says robots.txt changes can take up to 24 hours. Beyond that, I judge progress on four to eight weeks of runs.
Then count visits. Since May 2026, Google Analytics 4 has an AI Assistant default channel for sessions whose referrer "matches a list of AI Assistants." Look for Perplexity there, and treat it as a floor, since visits without a referrer land in Direct. Setup is in how to track AI traffic in GA4. Also ask new customers how they found you, since many buyers never click.
What doesn't work, or isn't proven
| Tactic | What the evidence says | What I'd do |
|---|---|---|
| Adding statistics and quotes for a promised lift | The GEO paper's Perplexity test used uploaded files, and its methods told the model to invent statistics and quotes. | Add real, sourced numbers because buyers want them. |
| llms.txt | Perplexity's crawler docs don't mention reading other sites' files. Ahrefs found 97% of 137,000 sites' files got no requests in May 2026. | Optional, low priority. See llms.txt and whether it works. |
| Joining the Publishers Program to get cited | The program shares ad revenue with publishers whose content is referenced. Nothing official says it changes which sources get picked. | Join for revenue if you publish, not for citations. |
| Getting onto a "trusted domains" list | Some guides say Perplexity boosts preset lists of authoritative domains. It documents source labels, which it doesn't link to source selection. | Add bylines and a corrections policy for readers. |
What the GEO paper's Perplexity test measured
The 2023 paper that introduced generative engine optimization (GEO) is where the "up to 40%" figure on agency pages comes from, and its Perplexity test gets cited as proof. Perplexity "does not allow the user to specify source URLs," the authors wrote, so they uploaded source texts as files and had Perplexity answer 200 queries from those files alone. Quotation Addition raised how much of the answer came from a source, by their measure, from 24.1 to 29.1.
That test skipped live retrieval, the part that decides whether Perplexity finds your page at all. And the paper's public code told the model to add statistics "even if hypothetical" and quotes "even though fake and artificial." The gains measured a reaction to invented evidence, so nobody should promise you a percentage lift from adding real numbers.
Tools that track Perplexity citations
I read each vendor's pricing page in October 2026. Some tools include Perplexity at the entry price, others count it as one of a few chosen models or keep it for Enterprise. Disclosure: I make InkieAI.
| Tool | Perplexity on the entry plan | Entry price a month |
|---|---|---|
| Otterly.AI | Included, 15 prompts | $29, or $25 billed annually |
| HubSpot AEO | Included, 25 prompts | $50, or $45 billed annually |
| Scrunch | Included, 125 prompts | $250 |
| Peec AI | One of 3 models you choose; extra models cost more | $95 |
| Profound | Enterprise only; the free trial excludes it | Custom |
| Writesonic | Not on Starter | $79, billed annually |
| InkieAI, which I make | Included on every plan, 30 questions checked weekly on Start | $99, or $79 billed annually |
Inkie also turns gaps into work: pages it writes after you approve them, exact edits, website fixes and outside sources to pursue. Its limits: weekly checks on Start where most trackers check daily, no Claude, Copilot, Grok, DeepSeek or Meta AI, and it's new, founded in 2026 and run by me alone from Stockholm. It can't guarantee a citation either. Compare more options in the best AI visibility tools.
Frequently asked questions
What is Perplexity SEO?
Perplexity SEO, also called Perplexity AI SEO or SEO for Perplexity, is the work of getting your pages cited and your brand named and recommended in Perplexity's answers. It overlaps with ordinary SEO: crawlable pages that answer real questions with specifics. The differences are Perplexity's own crawler and citations the buyer always sees. I compare the terms in GEO vs SEO vs AEO.
Is Perplexity good for SEO?
As a research tool it can help: Perplexity's Improve SEO strategy page pitches it for finding topics, competitor content and gaps. The page doesn't say where its search volume figures come from, so check them in a keyword tool. As a channel, every answer shows numbered sources, so a cited page is one click away.
Does Perplexity use Google?
Perplexity doesn't say so. It documents its own crawler, an index of "hundreds of billions of webpages" and partner crawlers, and names no search provider. Check PerplexityBot's access even if you rank well in Google.
How do I get my website cited by Perplexity?
Allow PerplexityBot and its IP ranges through robots.txt and your firewall. Publish one strong page per buyer question, answer first, with dated specifics. Earn accurate mentions on the threads, lists and review sites Perplexity already cites, and measure the share of runs that cite you over several weeks. No one can guarantee a citation.
Is SEO dead now with AI?
No. Perplexity answers from an index built by crawling, and a page PerplexityBot can't read contributes, at most, its headline and a short summary. Google's guide says optimizing for its AI search is "still SEO." What changes is that buyers see the sources, and many are other people's pages.
Run your first Perplexity check this week
Check your robots.txt and firewall for PerplexityBot today. Then pick the ten buyer questions closest to a sale, run each one five times in Perplexity, and fill in the sheet. The sources column will show whether your first job is access, your own pages or the outside sources Perplexity cites.
For a head start, the free AI visibility report checks 5 of your buyer questions across Perplexity, ChatGPT, Gemini, Google AI Overviews and AI Mode, and shows where you're named, recommended or cited and who shows up instead. No card needed.