Perplexity Visibility
Perplexity shows its sources, so being worth citing is the whole job
Perplexity attaches numbered sources to almost every claim it makes. That makes it the most legible of the answer engines, and the clearest test of whether your page is a source that other people build on or just more content about a topic.
The premise of a citation-first engine
Every claim in the answer arrives with a footnote. Your job is to be the footnote, which means being the source of something rather than a restatement of it.
That single design decision makes Perplexity the most useful of the answer engines to work on, because it shows its working. You can read which domains your category actually gets cited from, on which kinds of question, and compare that against what you have published.
It also makes the failure obvious. If your best page is a well-written summary of what four other people published, there is nothing there to cite that is yours, and an engine that summarises for a living has no reason to summarise your summary.
What we can and cannot say
How source selection appears to work, and where our knowledge stops
The retrieval logic is not published, and we will not pretend to reverse-engineer it. What is observable is the shape of the output, and the shape is consistent enough to work with.
Answers draw on several different domains rather than several pages from one site. Recent material is favoured for anything time-sensitive. Specific, checkable statements are cited more readily than general ones. And passages, not whole pages, are what gets used, which is why a claim split across three paragraphs is harder to draw on than the same claim written once, completely, in one place.
That is a description of behaviour, not a mechanism, and it is worth being precise about the difference. We optimise against the observable behaviour and the published guidance from search engines about clear, original, people-first content. We do not sell a theory of the ranking function.
The practical consequence of source diversity is worth sitting with. If an answer wants five domains, your fifth article on the same topic is not competing with the other four sites, it is competing with your own. Depth in one narrow place beats breadth across a wide one, and that is close to the opposite of how most content calendars are built.
What makes a source
Four properties that separate a cited page from an ignored one
We apply these as a checklist to existing pages before writing anything new. Most content estates fail on the first and the third.
Primary rather than derivative
The page is the origin of the claim. Your own delivery data, a documented methodology, a survey you ran, an aggregate pattern only you can see from the work you do. A summary of research other people published has no reason to be cited when the engine can summarise the originals directly.
Checkable
A specific number, a stated range, a named method, a defined sample. Vague competence is unusable. "Projects like this typically run eight to fourteen weeks, based on the last forty we delivered" can be quoted. "We deliver quickly" cannot be quoted by anything that has to stand behind what it says.
Current, and visibly so
A publication date, a review date, and a short note on what changed. For anything that moves, an undated page is a risk a retriever does not need to take when a dated competitor exists. This costs almost nothing and it is missing from most sites we look at.
Self-contained at the passage level
The figure, the qualifier and the source sit in the same place, so the claim survives being lifted out of the page. A number in one paragraph, its caveat two paragraphs down and its source in a footnote is a claim that cannot be used safely, however good the article is as a whole.
Where citations get earned
The question shapes that produce a source list
People use citation-first engines for research and due diligence more than for chat. These are the shapes of question that produce a list of sources, and therefore the shapes worth having a page for.
We build a fixed set of these for your category, run them regularly, and record which domains are cited rather than only whether you appear.
Questions that expect evidence
The highest-value group. A specific number with a stated method is what gets pulled in here.
- what percentage of [audience] actually [behaviour]
- is it true that [claim]
- what does the research say about [topic]
- sources for [claim] with dates
Cost, timeline and scope research
Buyers checking expectations before they contact anyone. Published ranges get cited; hidden pricing does not.
- typical cost of [service] in [country]
- how long does [project type] usually take
- what is included in a [service] engagement
Comparison and due diligence
Where a documented methodology or an honest trade-off page earns its place in the list.
- compare [approach] and [approach] with pros and cons
- what goes wrong with [project type]
- alternatives to [product] and why teams switch
- what should i check before hiring a [service]
Definitions and mechanisms
Commodity territory unless you can explain the mechanism better than anyone currently does.
- how does [process] actually work
- what is the difference between [term] and [term]
- explain [concept] with a worked example
These are examples of how customers in this market search, drawn from keyword research and from the questions that come up on sales calls. They are illustrative, not a volume claim — the actual demand in your area is something we size before recommending anything.
Questions
What people ask about citation-first search
If Perplexity answers the question, why would anyone visit our site?
Some will not, and that is a real cost worth naming. But the click behaviour on citation-first engines is different from a search results page, because the citation arrives attached to a claim the reader has already accepted. People follow it to check something, to go deeper, or because they now want to talk to whoever produced the number.
The important variable is what you sell. If your product is the article, a citation is a partial substitute for the visit. If your product is a service and the article was marketing, a citation is a recommendation with a footnote, and the trade is heavily in your favour.
We have no original data. Are we out of the running?
No, but you are competing on a harder footing, because a summary of published material is exactly what an answer engine already does for itself. If everything on your site is a restatement, there is nothing to cite that is specifically yours.
Most businesses have more original material than they think. Aggregate patterns from your own delivery work, the failure modes you see repeatedly, a documented methodology, price ranges by project type, timelines by scope. None of that requires a research budget. It requires being willing to publish something specific enough to be wrong about.
Should we allow or block PerplexityBot?
It depends entirely on what your content is for. Blocking reduces the chance of your material being used without a visit, and it also removes you from consideration as a source. There is no configuration that gets you the citation without the crawl.
For a business whose content exists to win clients, allowing access is usually right, because absence is worth less than an unattributed paraphrase. For a publisher whose product is the content itself, the calculation is genuinely different and reasonable people land differently. We will work through it with you and write the decision down, rather than applying a default.
What actually makes a page citable?
Four things, in our experience. It contains something specific enough to be checked, ideally a number or a stated method. The claim is self-contained, so the figure, its qualifier and its source sit together rather than being spread across the page. It carries a visible date. And it is the origin of the claim rather than a summary of somebody else being the origin.
None of that is exotic and all of it makes the page better for a human reader, which is the test we apply before making any change.
Can you guarantee we will show up in the source list?
No. Selection happens inside a system nobody outside it controls, the same question can produce different sources on consecutive runs, and we do not claim to know the ranking logic.
We work on eligibility and quality: reachable, current, specific, checkable, and the origin of something rather than a summary of it. That is the honest offer, and if an agency offers you more than that here, the extra is invented.
Find out what your category actually gets cited for
We will run a set of research questions from your market and bring you the source lists rather than the answers. It is usually clear within ten minutes what the cited domains have that you do not.
Related
Where to go next
Last updated · Reviewed by Zubair Afzal