Perplexity SEO 2026: How to Get Cited in Perplexity Answers

Ask Perplexity a question and the answer comes back with a short row of numbered sources under it. Getting your page into that row is what people mean by Perplexity SEO. This guide covers how Perplexity builds an answer, which crawlers you have to let in, what our own tracker counted, and what the research says about pages that get quoted.

Perplexity SEO in Brief: How Perplexity Picks Sources

Perplexity SEO is the work of getting your pages cited in Perplexity’s answers, and it starts with being reachable: Perplexity’s Help Center says that when you ask a question, “it uses advanced AI to search the internet in real-time,” then summarizes what it finds, and “Each answer includes numbered citations linking to the original sources” (Perplexity Help Center). In our own tracker, Perplexity cited exactly 10 sources per prompt. So how does a page earn one of those 10 seats?

Perplexity doesn’t publish a list of ranking factors. What it does publish is narrower and more useful: which crawler finds pages, how robots.txt is handled, how the firewall should be set up, and what kinds of sources it looks for (“articles, websites, and journals,” in the Help Center’s words). Everything else in this guide comes from two places. One is the 2023 research paper that coined generative engine optimization, which tested its methods on Perplexity. The other is our own tracking data from October 21, 2026.

Who is this for? Site owners, bloggers and marketers who want their pages to show up as sources in Perplexity. If you searched for Perplexity AI SEO, it’s the same thing: the product is Perplexity, the company calls it an AI-powered search engine, and the goal is a citation rather than a blue link.

Key Takeaways

Perplexity cited exactly 10 sources on each of the 5 prompts we tracked on October 21, 2026 (Source: our Otterly.AI test). Those seats go to pages that PerplexityBot can reach, so start with robots.txt and your firewall. After that, the strongest tested lever on Perplexity is a page that quotes named sources and backs claims with data.

How a Perplexity Answer Is Built

A Perplexity answer is built in four steps that the Help Center names in order: understanding your question, searching the web, summarizing the information, and citing sources (Perplexity Help Center). Its Research mode goes further: it “performs dozens of searches automatically, reads hundreds of sources,” and delivers a report “in just 2-4 minutes.” Why does that matter for your site? Because the mode a reader picks changes how many sources an answer can carry.

Perplexity also calls itself an answer engine. Its Help Center defines that as “a tool designed to give you direct, detailed answers to your questions,” and adds a line worth repeating to anyone who relies on it: “we encourage you to double-check sources for added confidence” (Perplexity Help Center).

Quick answers, Pro Search and Research

Perplexity’s Pro Search page compares the two everyday modes side by side (Perplexity Help Center), and its plans page lists how often each one is available on the free plan (Perplexity Help Center):

Mode	How Perplexity describes the sources	Free plan access
Standard search	“Surface-level, 1–2 sources”	Practically unlimited basic searches
Pro Search	“Comprehensive, dozens of sources”	3 per day
Research	Dozens of searches, hundreds of sources	1 per month

Plan matters too. The plans page says Education Pro adds “10x as many citations in answers.” Put those facts together and one thing is clear: there’s no single “Perplexity result” for a question. A page can sit among dozens of sources in a Pro Search answer and still miss the shorter list in a quick answer.

Pro Search also shows its work. The Help Center says you see “not just the summary, but also how the AI broke down your question and approached the research.” Open that breakdown on a question your customers ask. The sub-questions it lists are a good map of what your page should answer.

Where the sources come from

Most sources come from Perplexity’s own search index. Its robots.txt article says “PerplexityBot indexes pages similarly to other search engines,” and that “Perplexity partners with third-party crawlers to help build our search index” (Perplexity Help Center). The web isn’t the only input, though:

• Premium data sources. Perplexity lists Wiley, PitchBook Essentials, CB Insights and Statista, which users can pick from the sources menu or call with @statista and similar tags (Perplexity Help Center).
• The reader’s own files. Uploaded documents and, in organizations, internal files can be chosen as sources instead of the web, or together with it.
• The model. Pro users can pick the model that writes the answer, from OpenAI’s GPT-5 to Anthropic’s Claude Sonnet 4.6 and Google’s Gemini 3.1 Pro. The model writes; the search step decides which pages it gets to read.

Here’s the catch for publishers. If your page only repeats market numbers that a reader can now pull straight from Statista inside Perplexity, your page adds little. That’s our reading, not a Perplexity rule, but it points the same way as the research later in this guide: original facts beat repackaged ones. A user-side explainer of how the Perplexity search engine works is on our list of upcoming guides.

What Our Tracker Saw: 10 Sources on Every Prompt

Perplexity cited exactly 10 sources on every one of our 5 prompts, 50 citations in total, when Otterly.AI checked them on October 21, 2026, and none of those citations pointed to our site. We tracked a small marketing website our own team runs, with the market set to the United States and the language set to English. What does a flat 10 tell you?

Some setup details first, because they shape the numbers. Otterly.AI’s trial lasts 7 days and needs no card. Its onboarding wouldn’t continue with fewer than 15 prompts, so we kept our 5 prompts plus 10 of its suggestions and only counted our own 5. The first report took several minutes; the pages said “Today’s data is still processing” until the results arrived.

The five prompts were about our own niche: what GEO is, how to get cited by ChatGPT, tools to track AI visibility, optimizing for AI Overviews, and AI SEO agencies in Taiwan. Otterly reported 0% brand coverage on every prompt and 0 citations of our domain. Across all four engines, the most cited page on Otterly’s Citations page was Google’s guide to optimizing for generative AI features, cited 7 times.

Why exactly 10? We can’t see inside Perplexity, and one day of data can’t settle it. Two facts are worth knowing, though. Otterly reads one view of each engine, which may not match what a signed-in user gets in Pro Search, where the Help Center talks about “dozens of sources.” And Perplexity’s own Search API, covered later in this guide, returns 10 results by default. That may be a coincidence. Either way, the practical lesson is the same: in the answers we tracked, the number of seats was fixed, so a new source only gets in by pushing another one out.

Three cautions keep this honest. It’s one day, not a trend. The prompts are in one niche. And the counts are Otterly’s view of Perplexity, not every reader’s.

Perplexity vs ChatGPT, AI Overviews and Copilot: Citation Counts Compared

Perplexity was the only engine whose citation count never moved: 10 on every prompt, while ChatGPT ranged from 17 to 37, Google AI Overview from 2 to 11 and Microsoft Copilot from 5 to 8. Perplexity produced 50 of the 244 citations Otterly.AI counted that day, about 20%. Does a steady 10 make Perplexity easier or harder to win?

Prompt	Perplexity	ChatGPT	Google AI Overview	Microsoft Copilot
What is GEO?	10	37	9	5
How to get cited by ChatGPT	10	22	11	5
Best tools to track AI visibility	10	29	7	5
Optimizing for AI Overviews	10	17	7	5
Best AI SEO agencies in Taiwan	10	25	2	8
All five prompts	50	130	36	28

Source: Indie Stack Lab tracking in Otterly.AI, October 21, 2026, United States, English.

Both, in different ways. ChatGPT cited more pages per answer, so there were more seats, but each one got a smaller share of attention. AI Overview leaned on Google’s own index and dropped to 2 sources on one prompt. Perplexity sat in the middle with a fixed list. If you want to compare the other engines in depth, our guide to how ChatGPT search decides which pages to cite and our explainer on which pages Google AI Overviews lean on use the same tracking data.

What doesn’t change between engines? The basics. Each one needs its own crawler let in, and each one quotes pages that answer the question plainly.

Let PerplexityBot and Perplexity-User Crawl Your Site

To appear in Perplexity’s answers, allow PerplexityBot: Perplexity’s crawler documentation says “To ensure your site appears in search results, we recommend allowing PerplexityBot in your site’s robots.txt file and permitting requests from our published IP ranges,” and that changes “may take up to 24 hours” to take effect (Perplexity docs). Perplexity runs two agents, and they follow different rules. Which one do you actually control?

User agent	What it does	robots.txt	Published IP list
PerplexityBot	Finds and links websites in Perplexity’s search results; “not used to crawl content for AI foundation models”	Followed	perplexitybot.json, 8 IPv4 ranges
Perplexity-User	Visits a page when a user’s question needs it, and may link it in the answer	“generally ignores robots.txt rules,” since a user asked for the page	perplexity-user.json, 4 IPv4 ranges

Range counts are as we read the two files on October 24, 2026. Perplexity says to always use the current lists, because “These addresses are updated regularly.”

A robots.txt that lets Perplexity in

If your robots.txt has no rule for these agents, they’re already allowed. If it has a blanket block, or you want to be explicit, add:

User-agent: PerplexityBot
Allow: /

User-agent: Perplexity-User
Allow: /

The full user-agent strings, per the same docs, contain PerplexityBot/1.0 and Perplexity-User/1.0. Use those when you filter server logs.

What blocking does and doesn’t do

Blocking PerplexityBot keeps your text out of its index, but not your name. The robots.txt article says PerplexityBot “will not index the full or partial text content of any site that disallows it via robots.txt. However, if a page is blocked, we may still index the domain, headline, and a brief factual summary” (Perplexity Help Center).

Worried about AI training? That’s a separate question. The same article says “Perplexity does not build foundation models, so your content will not be used for AI model pre-training.” So unlike some other AI companies, there is no training crawler to block separately.

Perplexity also addresses the old complaint that its crawlers ignored robots.txt. It says users could once prompt it to summarize a blocked URL, and “this feature has been disabled to prevent misuse.” It adds that it has “updated our agreements to ensure these providers also respect robots.txt,” meaning the third-party crawlers that help build its index.

Perplexity-User is the exception. Because it fetches pages for a person, a robots.txt line won’t reliably stop it. If you truly want it out, that takes a firewall rule. Think twice before you write one, though: it also stops Perplexity from opening your page when a reader asks about it by name.

Check Your Firewall and Your Audit Tool

A robots.txt “Allow” means nothing if your firewall turns the bot away: Perplexity’s crawler docs say sites behind a web application firewall “may need to explicitly whitelist Perplexity’s bots,” and give steps for 2 providers, Cloudflare and AWS (Perplexity docs). Bot protection, rate limits and “are you human” checks can all block a crawler without anyone noticing. How would you know?

Firewall rules Perplexity documents

For Cloudflare, the docs describe a custom rule under Security, then WAF, then Custom rules:

1. Field User Agent, operator Contains, value PerplexityBot or Perplexity-User.
2. And field IP Source Address, operator Is in, value: the ranges from the two official JSON files.
3. Action: Allow, so the requests skip your other security rules.

For AWS WAF, the steps are to create IP sets from the same JSON files, add string match conditions for the two user agents, combine them into allow rules, and attach those rules to your Web ACL with “higher priority than blocking rules.” Perplexity’s tip is to match both the user agent and the IP address “for enhanced security”; in plain terms, a scraper that copies the bot’s name still won’t come from Perplexity’s addresses. It also suggests fetching the IP lists on a schedule, since they change.

What our audit tool showed

We checked our own site with a free Semrush account, and it didn’t settle the Perplexity question. The Site Audit overview showed a “Blocked from AI Search” panel that named ChatGPT-User, OAI-SearchBot, Googlebot and Google-Extended. PerplexityBot wasn’t one of the names on that overview, and we didn’t open the panel’s detail view. The same audit gave our site an AI Search Health (beta) score of 95%, and Perplexity still cited us 0 times. For the rest of that report, see our write-up on what else a free 100-page Semrush audit flagged on our site.

So don’t take an audit dashboard as proof that Perplexity can get in. Check the server logs instead. A quick look on a typical Linux host:

grep -i "perplexitybot" /var/log/nginx/access.log | awk '{print $9}' | sort | uniq -c

That counts the status codes PerplexityBot received. Lots of 200s means it’s reading your pages. 403s, 429s or challenge pages mean something in front of your site is blocking it. No lines at all can mean it hasn’t visited yet, or that requests are stopped before they reach your server, which is when you check the CDN’s own logs.

Pages Perplexity Can Quote: Structure That Helps

The one controlled test of content changes on Perplexity is the 2023 GEO paper: on Perplexity.ai, adding quotations from credible sources did best, with a 22% gain over the baseline on the authors’ Position-Adjusted Word Count metric, while keyword stuffing did 10% worse than the baseline (Aggarwal et al., arXiv). The researchers also report that citing sources and adding statistics gave gains “of up to 9% and 37%” on their two metrics. Is a 2023 test still useful?

As a direction, yes, with limits. The team tested 200 queries, and because “perplexity.ai does not allow the user to specify source URLs,” they uploaded the source text as files rather than letting Perplexity search the web. Perplexity has changed a lot since then. The paper, accepted to KDD 2024, is still the only published study we know of that ran these page changes through Perplexity itself, and its lesson matches what Perplexity says it wants: answers “supported by citations from reputable news organizations, academic publications, and established content sources” (Perplexity Help Center).

For the wider picture, including how the researchers scored visibility, see our breakdown of the methods the GEO researchers tested and how they scored them.

A page structure that gives Perplexity something to quote

This list is our judgment, built from the paper and from the pages we saw cited:

• Answer under every heading. Put the direct answer in the first sentence after each heading. A Perplexity answer often lifts one sentence; make that sentence complete on its own.
• Name your sources in the text. Quote the official page or the named expert, and link to it. The paper’s best method on Perplexity was exactly this.
• Use numbers you can back up. A figure with a source and a date beats an adjective. First-hand numbers, like a test you ran, are even better, because nobody else has them.
• Write descriptive headings. “How long does a refund take?” tells a search step what the section answers. “Details” doesn’t.
• Keep facts in text. Prices, steps and specs inside images or behind tabs are harder for any crawler to read. Put them in plain HTML text or a table.
• Date your page. Show when it was last checked. Perplexity’s Help Center stresses “the most up-to-date information available.”
• Skip keyword stuffing. On Perplexity, the paper found it hurt.

What did the cited pages in our tracker have in common? The top one, Google’s own guide, is an official source that answers one question in depth. That’s hard to beat with a general blog post. It’s easier to beat with a page that adds something the official page can’t, such as first-hand results.

Check Where You Show Up With the Perplexity Search API

The Perplexity Search API lets you see which pages Perplexity’s search returns for a question: it returns “ranked web results as JSON, not a generated answer,” costs $5.00 per 1,000 successful requests, and returns 10 results by default (Perplexity docs). There’s also an interactive Playground that, per the same docs, needs no API key. Why bother, when you could just ask Perplexity?

Because it gives you a repeatable check. Typing questions into the app gives a different answer each time and depends on your mode and plan. The API gives ranked results with a title, URL, snippet and dates, which you can save and compare month to month.

A simple routine for a small site:

1. Write down 10 to 20 questions your customers really ask.
2. Run each one in the Playground, or through the API with max_results set anywhere from 1 to 20.
3. Note whether your URL appears and in what position.
4. Run the same questions with search_domain_filter set to your domain. If nothing comes back, Perplexity’s search may not have your pages at all, which points back to crawler access.
5. Use the country parameter for a market such as the US, and search_recency_filter if your topic moves fast.

Two limits apply. The domain filter takes up to 20 domains per request, in include or exclude mode, not both. And ranked results are not the same thing as citations in an answer: the answer step can still pick a different set. Treat the API as a proxy that tells you whether you’re in the running, not as a citation count. For cost, the docs also list Fast Search at $1.00 per 1,000 requests and a rate limit of 50 query units per second.

Perplexity Computer and SEO: What the Agent Changes

Perplexity Computer is an agent, not a search mode: the Help Center sums it up as “Perplexity answers your questions. Computer does your work,” and says Max subscribers get 10,000 credits a month to run it (Perplexity Help Center). If you searched for Perplexity Computer SEO, the short answer is that Perplexity documents no separate optimization for it. So what changes?

Here’s what the Help Center does say. Computer needs an active subscription and runs on the web, iOS and Android. It can run “parallel web searches,” monitor email and calendars, connect to apps such as Gmail, Slack and Notion, and chain steps together. Regular Ask searches “remain unlimited and don’t use credits.” Its listed capabilities start with “Search-native intelligence for accurate, cited information,” the same cited search the rest of this guide is about.

The example task on that page is the interesting part for site owners: “Research our top 5 competitors, compare their pricing, create a slide deck with the findings, and email it to the team.” An agent doing that reads pricing pages, feature pages and comparisons, then turns them into a table. Our take: if your prices and plan limits are only in images, behind a “contact us” form or out of date, an agent can’t put you in that table. Plain-text pricing and current specs are the Computer-era version of good SEO.

Tracking Perplexity Citations: Which Tools Include It

Of the four trackers we set up in October 2026, only Otterly.AI showed Perplexity results in its base product, with plans from $29/month for 15 prompts; SE Ranking showed a Perplexity widget, and Writesonic’s Starter trial left Perplexity out. Do you need a tracker at all? Not on day one, but you need some record, or you can’t tell whether a change worked.

Tool	Perplexity in what we saw	Notes from our trial
Otterly.AI	Included in the base product, next to ChatGPT, Google AI Overview and Microsoft Copilot	7-day trial with no card; 15 prompts minimum to finish setup; Lite $29/month for 15 tracked prompts
SE Ranking	Perplexity widget on the project overview, with AI Overview, AI Mode, ChatGPT and Gemini	14-day trial with no card; 20 tracked prompts; at least 1 prompt required to launch; results were still collecting when our log closed
Writesonic	Not on the Starter trial, which tracked ChatGPT, Gemini and Google AI Overviews	7-day trial of a paid plan with no card; Starter $79/month billed annually for 50 prompts
Semrush (free account)	Position Tracking offered Google, ChatGPT, Gemini and more as engines	AI prompt tracking sits in paid plans; the free account tracks 10 keywords

Source: Indie Stack Lab trials, October 21, 2026.

For full setups and prices, our roundup of which AI visibility trackers cover Perplexity and at what price compares all four. The Otterly and Writesonic results on the same prompts are in our write-up of how Otterly.AI and Writesonic handled our five prompts, the SE Ranking project is in our notes on what SE Ranking’s trial set up for AI prompts, and the three platforms Writesonic’s trial covered are in our rundown of what Writesonic’s Starter trial tracked on our site.

No budget? Track by hand. Once a week, run your 10 most important questions in Perplexity, save the list of sources, and note whether you’re in it. It takes about as long as a coffee break and gives you a baseline that costs nothing.

Perplexity SEO Mistakes to Avoid

The costliest Perplexity SEO mistake is blocking PerplexityBot by accident, through a firewall or a blanket robots.txt rule, and trusting an audit score instead of the logs: our own site scored 95% on Semrush’s AI Search Health and that panel never named PerplexityBot. The rest of this list is our judgment, drawn from the facts above. Which of these is your site making?

1. Blocking PerplexityBot to stop AI training. Perplexity says it doesn’t build foundation models and PerplexityBot isn’t used for training. Blocking it only removes you from answers.
2. Allowing it in robots.txt but challenging it at the CDN. Bot protection that serves “are you human” pages blocks the crawler just as well as a Disallow line.
3. Assuming your Google ranking carries over. Perplexity uses its own index and its own crawler. Check its search directly.
4. Changing content before you measure. Save a baseline of who gets cited for your questions first. Otherwise you can’t tell a win from noise.
5. Stuffing keywords. In the GEO paper’s Perplexity test, it scored 10% worse than doing nothing.
6. Reading one day as a trend. Our own numbers are a single day. Look at weeks, not a screenshot.
7. Hiding the facts. Prices, specs and steps in images or PDFs only are hard for crawlers and agents to quote.

Your First Week of Perplexity SEO: A Checklist

A first week of Perplexity SEO needs about five tasks, and none of them costs money: check access, fix blocks, measure, then improve a few pages. Perplexity’s crawler docs say robots.txt changes can take up to 24 hours, so do the access work first. Ready to start?

• Day 1: Read your robots.txt. Make sure PerplexityBot isn’t disallowed, directly or by a User-agent: * block.
• Day 1: Check your firewall or CDN. Add an allow rule that matches the user agent and the official IP lists.
• Day 2: Read your server logs. Look for PerplexityBot requests and the status codes they got.
• Day 3: Build a baseline. Run 10 customer questions in Perplexity or the Search API Playground and save who gets cited.
• Days 4 and 5: Improve two pages. Pick pages that answer those questions. Put the answer first, add sourced numbers and quotes, and move key facts out of images.
• After a month: Rerun the same questions. Compare against your baseline, not against a feeling.

If you also care about ChatGPT, most of this list carries over, and our guide to OpenAI’s crawler rules and what earns a ChatGPT citation covers the differences. Perplexity changes quickly, so check its Help Center and crawler docs before you rely on any single fact here, including ours.

References

• Perplexity Help Center: How does Perplexity work?
• Perplexity Help Center: What is Perplexity?
• Perplexity Help Center: What is an answer engine, and how does Perplexity work as one?
• Perplexity Help Center: What is Pro Search?
• Perplexity Help Center: Which Perplexity Subscription Plan is right for you?
• Perplexity Help Center: Premium Data Sources
• Perplexity Help Center: How does Perplexity follow robots.txt?
• Perplexity Help Center: What is Computer?
• Perplexity Help Center: Data Collection at Perplexity
• Perplexity docs: Perplexity Crawlers
• Perplexity docs: Search API Overview
• Perplexity: PerplexityBot IP ranges (JSON)
• Perplexity: Perplexity-User IP ranges (JSON)
• Aggarwal et al.: GEO: Generative Engine Optimization (arXiv 2311.09735, KDD 2024)