Perplexity is an AI answer engine that shows its sources as numbered links beside the answer. That one design choice makes it the easiest place for a real estate agent to find out whether AI tools are using their website.
Why Perplexity is worth testing first
Most AI visibility work is guesswork because you cannot see what the tool used. Perplexity removes that problem. Ask a question, read the answer, and look at the numbered list of sources next to it. The pages that shaped the answer are right there.
For an agent, that means a real test rather than a score from a tool that claims to measure AI visibility. You can ask exactly what a client would ask about your area and see whether your page is among the sources, or whether a portal and the local board answered it instead.
The visible citation also changes what a mention is worth. A reader who sees your page cited can click it in that moment, while they are researching. That is closer to a referral than to a search result.
What Perplexity documents, and what it does not
Being precise here matters, because there is a lot of confident advice about AI search that nobody can support.
Perplexity's crawler documentation describes two separate bots, and they behave differently.
PerplexityBot crawls and indexes pages so websites can be surfaced and linked in search results. Perplexity states it is not used to train AI foundation models. It publishes the IP ranges this bot uses at a JSON endpoint so publishers can verify the traffic is genuine.
Perplexity-User supports user actions. When someone asks a question that requires visiting a page, this agent fetches it so the answer can include a citation. Perplexity documents that because these visits are triggered by a person, this one generally ignores robots.txt rules. It has its own published IP range endpoint.
What Perplexity does not publish is the part everyone wants: how it decides which sources to rank above others for a given question. There is no documented ranking formula for citations. Anyone offering a guaranteed method for getting cited is describing a pattern they have observed, at best, and inventing one at worst.
So the honest position is this. The crawling and access rules are documented and checkable. The selection mechanics are not. We will separate the two throughout.
The ten-minute test
Do this before changing anything on your site, because it sets your baseline.
Write down five questions a real client would ask, in the words they would use. Not keywords. Questions like: how long are townhouses taking to sell in this neighbourhood, what should I know before buying in that building, is this area good for families with young kids, which agent knows this part of town, what are strata fees like in those buildings.
Ask each one in Perplexity, naming your neighbourhood and city explicitly.
Then record, for each question, which sources appear in the numbered list. Portals, your local real estate board, a news site, a competitor, a forum, or you.
That list is your honest starting position, and it is more useful than a percentage. If the board answers every statistical question and a competitor answers every lifestyle question, you now know what to write. If a forum thread from 2019 is being cited for your neighbourhood, that is an open gap.
Repeat monthly with the same questions. These answers vary between runs and over time, so a single result means little and a pattern across three months means something.
Why the portals usually win, and where they cannot
The recurring frustration is seeing a portal cited for a neighbourhood question when you have a page about that neighbourhood.
The reason is usually structural. The portal has a page dedicated to that one area, with the figures stated plainly, updated on a schedule. A typical agent area page opens with a photo gallery, a paragraph about the area's charm, and a listings widget. The answer to the question is either absent or buried.
Portals also have a real weakness, and it is the whole opportunity. They hold transaction data and nothing else. They cannot say what showings have been like there for the past six weeks, which two streets behave differently from the neighbourhood average, why one building trades below its neighbours, or what a buyer should check before making an offer on a unit in it.
That knowledge is yours and it is not in any dataset. Written as plain sentences on a page, it is exactly the kind of original, first-hand material Google's helpful content guidance describes, and it is the only content on the subject that a portal cannot reproduce. Our post on neighbourhood pages covers how to build one that holds this properly.
What a citable page looks like
The pattern that shows up repeatedly, across AI surfaces rather than Perplexity alone, is that pages answering one specific question directly get quoted and pages that circle a topic do not.
Answer in the first two sentences under the heading. If the page is about how long homes take to sell in an area, the time should appear immediately, with the period it covers and where the figure came from. Then the explanation underneath.
One question per page. A page trying to cover buying, selling, schools, and restaurants in one area gives an assistant nothing clean to pull. A page about one question can be quoted in a sentence.
Write the facts in text. An infographic containing your key figures is invisible to anything reading the page, because a machine sees an image file. Whatever matters has to exist in the words.
Use headings that match how people ask. A heading reading "How long do townhouses take to sell in this neighbourhood" is doing more work than one reading "Market insights".
Attribute your facts. Say which board or agency published a figure and for which month. A page that names its sources is easier to trust and easier to quote, for a reader and for a machine.
Date the page honestly. Show when it was published and when it was last updated, and only change the update date when you actually changed something.
Our post on what AEO means for real estate covers the structural side of this in more depth.
The access question, and why blocking is usually wrong
Some agents ask how to keep AI tools off their site. Before doing anything, understand that the two bots respond differently, so one rule does not cover both. PerplexityBot follows robots.txt. Perplexity-User, by Perplexity's own documentation, generally does not, because a person asked for that page.
For almost every agent, blocking is the wrong instinct anyway. You are not a publisher whose revenue comes from people landing on your article. You are a business that benefits from being named when somebody asks who can help them. Being absent from an answer is not a defensive position.
If you want to verify what is actually visiting, check your server logs for the documented user agent strings and confirm the requests against the published IP ranges. Verification matters because user agent strings are easy to copy, and traffic claiming to be a known crawler is not always that crawler.
How this fits with everything else
Perplexity is one surface. The work that gets you cited there is mostly the same work that helps on other AI surfaces and in ordinary search, which is convenient, because you cannot run separate strategies for each one.
What that shared work looks like: pages built around real questions, answers stated plainly near the top, local facts only a working agent holds, sources named, dates accurate, and content in text that anything can read. Our post on getting leads from AI search covers the wider picture, including what happens after someone arrives from one of these answers.
What it does not look like is stuffing a page with phrases about AI, publishing a weekly post nobody needs, or paying for a visibility score. None of those change whether a page answers a question better than the alternatives.
Being honest about what you cannot control
Three limits worth stating plainly, because the alternative is disappointment.
Results vary. The same question asked twice can produce different sources, and location and account history affect what appears. Track patterns over months rather than reacting to single runs.
Selection is undocumented. Perplexity publishes how its crawlers behave and does not publish how sources are chosen. Any specific claim about the ranking mechanics is unverified, including ones made confidently.
Coverage is uneven. In markets where a board publishes detailed public data and a news outlet covers housing closely, those sources are hard to displace on statistical questions. The winnable ground is the local, experiential question where nobody has published a good answer, and that is where an agent has a genuine advantage.
The takeaway
Perplexity shows its sources, so you can stop guessing about AI visibility and just check. Ask five real client questions about your area, note who gets cited, and repeat monthly. Then write the pages that answer the questions nobody covered well, with the answer near the top, in plain text, attributed and dated. The crawler rules are documented and worth understanding. The citation mechanics are not published, so treat anyone promising guaranteed placement as guessing.
