Perplexity is the answer engine people forget to optimise for.

Everyone obsesses over ChatGPT because of the raw user numbers. Fair enough. But Perplexity is where a specific, valuable kind of user goes: someone researching a decision, who wants sources, who clicks. If you sell anything considered, those are the people you want.

The good news is that Perplexity is the most transparent of the big engines about where its answers come from. It shows its citations right there, numbered, next to the text. That means you can actually reverse engineer it, which you cannot really do with a model that answers from memory and hides its reasoning.

Before I go further: I build LLM Pulse, a platform that tracks brand visibility across AI engines, and Perplexity is one of the models we monitor by default. So I have a horse in this race. I will try to be straight with you anyway, including about where other tools do a job better.

How Perplexity actually answers

It helps to understand the machine before you try to game it.

When you ask Perplexity something, it does not answer from the model's memory the way a raw chatbot does. It runs a search, pulls a handful of live web pages, reads them, and writes an answer grounded in those pages. Then it cites them. This is retrieval augmented generation, and the important word is retrieval.

So there are really two gates. First, your page has to be retrieved: it has to show up in the set of results Perplexity pulls for that query. Second, your page has to be worth quoting: once it is in the candidate set, the model has to decide your page is the best source for a given claim.

Most people only think about the second gate. The first one is where a lot of brands quietly lose.

If you are not retrievable, nothing else matters.

Perplexity leans heavily on its own index and on live search. It also draws on high-authority reference sources for general knowledge. What this means in practice is that classic search visibility still counts for a lot. If you already rank well for a topic and you have pages that clearly answer real questions, you are most of the way there. GEO did not delete SEO. It sits on top of it.

What kind of content earns citations

I have looked at a lot of Perplexity answers, both in our own data and by hand. A few patterns show up again and again.

It cites pages that answer the specific question directly. Not pages that dance around it. If someone asks for the best CRM for a two-person agency, the page that gets cited is the one with a clear, scannable comparison, not the 3,000-word thought piece that mentions CRMs in passing.

It likes structure. Clear headings that match the question. Short paragraphs. Lists and tables where a list or table is the honest format for the information. A model extracting a claim finds it faster in a clean table than in a dense block of prose, and it seems to reward that.

It likes specifics. Numbers, dates, named examples, prices. Vague marketing copy gives a model nothing quotable. A sentence like "affordable plans" is useless to it. "Plans start at 49 per month" is a fact it can lift and attribute to you.

It likes freshness on anything that moves. For queries where recency matters, and there are more of those than you think, Perplexity visibly favours pages that are recent or recently updated. If your comparison page still says "as of 2024", you are handing the citation to whoever updated theirs last week.

And it likes sources that are not you. More on that next, because it is the part people least want to hear.

Third-party sites do a lot of the work

Here is the uncomfortable truth. A large share of the citations that decide how your brand looks in Perplexity are on pages you do not own.

Ask Perplexity "is [your product] any good" and it will often pull a Reddit thread, a review roundup, a comparison article on some industry blog, a G2 page. Your own site might get cited too, but the model treats your own claims about yourself with the skepticism they deserve. When it wants to tell the user whether you are actually good, it goes looking for someone other than you saying so.

This changes the job. Getting cited in Perplexity is partly on-site work and partly the old, unglamorous work of being genuinely present in your category: mentioned in the roundups, discussed in the communities, reviewed on the platforms people trust, written about by people who are not on your payroll.

You cannot fake this at scale, and you should not try. But you can be systematic. Find the pages that already rank for your category's key questions, the ones Perplexity is likely pulling from, and make sure your product is accurately represented there. Sometimes that means a correction. Sometimes it means earning a mention that is not there yet. Reddit and other community discussion carry real weight here, which is uncomfortable for brands used to controlling the message, but it is the reality.

The blunt version: Perplexity mostly repeats what the rest of the web already says about you. If that consensus is thin or wrong, fix the consensus, not just your homepage.

The technical bit: let PerplexityBot in

None of the above works if you block the crawler.

Perplexity fetches pages with its own user agents. PerplexityBot is the one that crawls and indexes content for search. There is also a separate user agent it uses to fetch a page live when a user's question requires it in the moment. If your robots.txt or your firewall blocks these, you are invisible to the engine no matter how good your content is.

Check your robots.txt. Check whatever bot-management or WAF layer sits in front of your site, because that is where blocks happen silently, often added by a security team that never told marketing. Plenty of brands are blocking AI crawlers by accident right now and cannot work out why they never appear.

There is a real debate here about whether you want AI engines training on or scraping your content for free. That is a legitimate business decision. Just make it on purpose. Do not discover six months later that a default WAF rule quietly cost you every AI citation you might have had.

Beyond access, the usual technical hygiene applies. Server-rendered content the crawler can read without executing a pile of JavaScript. Clean, semantic HTML. Fast pages. A sensible sitemap. Structured data where it fits. None of this is new. It is just SEO fundamentals, and they still pay.

How to actually measure this

This is where most people are flying blind, and it is the part I care about most.

You cannot improve what you cannot see. "I asked Perplexity and I showed up" is not measurement. You asked once, from your account, with your history, on one phrasing of one question. That tells you almost nothing. Answers vary by prompt wording, by user, by day, and by how fresh the web is that morning.

Real measurement means tracking a stable set of prompts that matter to your business, across time, and recording three things: whether you get mentioned at all, whether you get cited with a link, and which sources Perplexity pulled instead of or alongside you.

That third one is the gold. When you can see the exact pages Perplexity cited for a question you care about, you stop guessing. You know which Reddit thread, which competitor's comparison, which review site is shaping the answer. That is your to-do list, handed to you.

This is what we built LLM Pulse to do. Perplexity is one of five core models we track on every plan, alongside ChatGPT, Gemini, Google AI Mode and Google AI Overviews, and we surface the citations and sources behind the answers, not just a mention count. You connect GA4 or Plausible and you can also see the actual visits and conversions Perplexity sends you, which closes the loop between "we got cited" and "it mattered". Pricing is public and starts at 49 per month, every plan has unlimited seats, and there is a 14-day trial and a hosted MCP server so you can pull your data straight into Claude or ChatGPT.

I am biased, so let me be useful instead of promotional.

If Perplexity is genuinely the only engine you care about and you want the cheapest honest way in, look at Otterly.ai. It starts at 29 per month, shows screenshots of real AI outputs, and does citation and link tracking well. It is a good tool and I will happily say so.

If you are an agency or a bigger team wanting strong regional and multi-language tracking with per-country competitor comparison, Peec AI is worth a look. Just check the model add-on pricing, because extra engines are billed per model.

And if you are a Fortune 500 brand with a procurement process and a budget to match, Profound has the deepest data set of anyone. It is enterprise, demo-led, no public pricing, and for that buyer it earns the money.

Where I think LLM Pulse wins is the middle: public pricing, five models included without an enterprise call, real traffic analytics rather than just mentions, and full white-label if you are running this for clients. Pick the tool that fits your actual situation. Just do not fly blind.

The honest summary

Getting into Perplexity is less mysterious than the GEO hype wants you to believe.

Be retrievable, which is mostly SEO you already know. Write pages that answer the exact question with specifics a model can quote. Keep the stuff that moves genuinely fresh. Earn accurate third-party coverage, because Perplexity trusts other people's words about you more than your own. Let the crawler in. Then measure it properly, over time, watching the sources, so you are editing reality instead of arguing with it.

Do that and you stop wondering whether you are in Perplexity. You know, you know why, and you know what to fix next.

The brands that win here are not the loudest. They are the ones the rest of the web already agrees are worth citing.