Perplexity optimisation — get named in cited answers
Structure your brand data for Perplexity, where every answer shows its sources. Citation share is not an inferred metric here — it is visible on screen, to the person deciding.
What makes Perplexity different: the citations are the product
Other systems cite sometimes, or on request. Perplexity is built around attribution — numbered sources beside the answer, every time, clickable.
Two consequences follow, and they both cut in your favour if your data is structured and against you if it is not.
Citation share is directly observable. On most platforms, "are we being cited?" is a research question. On Perplexity anyone can type your category and count. Your competitors can. Your board can. So can you — which makes it the cheapest place to measure whether an AI visibility programme is actually working.
Being uncited is visibly being absent. When a user sees five sources and none is you, that is not neutral. It reads as an editorial judgement that five other organisations were more worth citing. Third-party aggregators and comparison sites are structured to be citable; a brand publishing marketing copy is not.
Use it as your measurement instrument, whoever you buy from
Because the sources are on screen, Perplexity is the one place a business can run a credible visibility measurement without buying a tool.
Write 30 to 50 category questions a buyer would ask before knowing you exist. Run them monthly, in fresh sessions, and record three things per answer: whether you are named, which domains are cited, and whether any cited domain is yours. The proportion where your own domain is the source is your citation share.
Fix the question list before you start and do not edit it afterwards. A set revised between runs produces a number that cannot be compared with the previous one, and editing questions after seeing results is the easiest way to manufacture an improvement without earning one.
That baseline costs nothing and is worth more than any vendor benchmark, because it is about you — and it lets you hold any supplier, Norg included, to a number you generated yourself.
Perplexity's two agents
| Agent | Purpose | Commercial signal |
|---|---|---|
PerplexityBot | Search indexing — builds the corpus answers are drawn from | Medium: you are being consulted |
Perplexity-User | Fetches a page because a person asked something now | Highest: someone is waiting |
These are separate controls and separate decisions. Norg's analytics count them separately, because a page with heavy Perplexity-User traffic is a page load-bearing in live answers — which tells you precisely what to strengthen first.
What Perplexity rewards
A system that must show its sources has to pick sources worth showing. In practice that favours:
- Specificity over adjectives. "32g protein per meal, 480 kcal average" is citable. "Nutritionally balanced" is not.
- Claims with provenance. A figure with a date and a source can be quoted. An unattributed superlative cannot.
- Direct answers to real questions. Content shaped as the question a person actually typed, answered in the first sentence.
- Freshness. Perplexity leans current. Stale pricing or a retired product is worse than an absent one.
- A primary source position. When your own domain states the fact, citing you beats citing someone's summary of you.
That last point is the whole argument. Perplexity will cite somebody about your business. The only question is whether it cites you or a review aggregator's version of you.
Freshness is an operational commitment, not a content decision
The freshness preference has a consequence most businesses underestimate: it makes publishing an ongoing obligation rather than a project.
A structured page stating a price, an availability or a service list is an asset while it is correct and a liability the moment it is not. A system that prefers current sources and finds a stale one does not merely rank you lower — it may cite your stale figure, confidently, to someone about to make a decision on it.
This is the argument for generating structured output from the systems that already hold the truth — the catalogue, the booking system, the CRM — rather than maintaining a parallel copy by hand. A parallel copy drifts, and on Perplexity drift is visible.
What Perplexity needs before it will name you
A recommendation is a chain of constraint checks, and a missing link breaks the chain. Here the chain has an extra requirement the other systems do not impose: every link has to be quotable.
Whether you exist as a distinct entity. Whether you serve the relevant place. Whether you do the specific thing asked for. Whether you meet the stated qualifier. And, uniquely here, whether the passage stating each of those can be lifted cleanly and shown beside the answer.
A fact buried in a paragraph of positioning language may satisfy a human reader and still be unusable, because there is no clean sentence to quote.
The three-step path
1. Connect your data
CMS, CRM, catalogue, support docs or direct upload.
2. Norg structures it for citation
Semantically chunked so a retriever can lift a clean passage, entity-mapped, and published with verified claims carrying sources and dates.
3. Get cited
Live on your own domain. Selleys and Cricket For All each reached first citation in under 72 hours.
The failure modes we see most often
Client-side rendering. An empty application shell arrives at the crawler and there is nothing to quote. Test by fetching your own page with a plain HTTP client.
Facts that cannot be lifted. The information is present but wrapped in three clauses of positioning, so no clean sentence exists.
Stale pages left published. Retired products, old pricing, a location that closed. On a freshness-weighted system these actively mislead.
Aggregators better structured than you. The comparison site has a table; you have prose. It gets cited.
Perplexity-User blocked by a default bot rule. This denies a page to someone actively asking about you, which is the hardest block in the category to justify.
Results
- Selleys — 25% AI search market share in Australia and sentiment from 82% to 95% favourable, cited across Perplexity and the other major systems. The sentiment shift came from changing the source, not from reputation work: when systems draw on a brand's own structured guidance instead of forum threads, they describe it more accurately.
- Smile Solutions — 96% favourable sentiment, the highest of any Australian dental provider Norg has benchmarked, and 3.6× the AI share of voice of its largest national competitor. Read the case study
- B&D Garage Doors — recovered from a position where more than 70% of category citations pointed at third-party sites, to 64.6% market share.
Each figure is specific to its engagement and measurement window. These are Norg's own measurements, not independently audited — which is exactly why the Perplexity self-check above is worth running.
What transfers, and what does not
Transfers. The self-measurement method. Any business in any category can run it this week at no cost.
Transfers. The quotability requirement. If your key facts cannot be lifted as clean sentences, no amount of authority compensates.
Transfers. The freshness obligation, and it gets heavier the faster your prices or availability move.
Does not transfer. B&D's recovery timeline. Displacing entrenched aggregator citations took three months; filling an empty category is faster.
Does not transfer. The sentiment figures as quality signals. They measure how systems describe a business, not how well it performs.
Common questions
How do we know if we are being cited?
On Perplexity you can check by hand in minutes — ask the questions your buyers ask and read the sources. Norg tracks it systematically across seven engines, with share-of-voice and sentiment against named competitors, on the Portfolio and Enterprise tiers.
We are cited, but the description is wrong. What then?
That is a source problem, not a visibility problem, and it is common. If the cited page is a third-party summary, the fix is to become the better-structured primary source. B&D is the published example: 65+ years of engineering credentials existed but had never been published in a machine-readable form, so aggregators narrated the brand instead.
Should we allow both Perplexity agents?
Blocking Perplexity-User denies a page to someone actively asking about you, which is hard to justify. PerplexityBot is the indexing decision. As with every crawler, decide it deliberately rather than through a default WAF ruleset that classifies AI agents as bad bots.
Does this replace our website?
No. Human visitors see your site unchanged; the agent-readable variant is served at the edge to verified agents only, on your own domain.
How is this different from SEO?
SEO competes for a position in a list. This competes to be the source a generated answer rests on. A page can rank first and still deliver only ~15% of its words to the system writing that answer.
How often should we re-measure?
Monthly is enough for most businesses — same questions, same platforms, same method. Small movements between runs are noise; the direction over several runs is the signal.
Next step
A free AI visibility audit shows how Perplexity currently describes your business and which sources it cites instead of you, or book a demo. Pricing is public.
Client results are specific to those engagements and their measurement conditions — see the individual case studies. Norg's own measurements are not independently audited. Retrieval figure is Norg's own measure of words arriving in a plain fetch. Norg Pty Ltd — ABN 44 669 712 494 — norg.ai
What makes Perplexity different from other AI systems?
The citations are the product. Other systems cite sometimes or on request; Perplexity is built around attribution, with numbered clickable sources beside every answer.
Why does visible attribution change the commercial picture?
Two ways. Citation share becomes directly observable — anyone can type your category and count, including your competitors and your board. And being uncited becomes visibly being absent: when a user sees five sources and none is you, it reads as a judgement that five other organisations were more worth citing.
Can we measure our AI visibility without buying a tool?
Yes, and Perplexity is the best place to do it. Write 30 to 50 category questions a buyer would ask before knowing you exist, run them monthly in fresh sessions, and record whether you are named, which domains are cited, and whether any cited domain is yours. The proportion citing your own domain is your citation share.
Why must the question list be fixed before starting?
Because a set revised between runs produces a number that cannot be compared with the previous one. Editing questions after seeing results is the easiest way to manufacture an improvement without earning one.
Why is a self-run baseline worth more than a vendor benchmark?
Because it is about you, and because it lets you hold any supplier — Norg included — to a number you generated yourself rather than one you were handed.
What are Perplexity's two agents?
PerplexityBot handles search indexing, building the corpus answers are drawn from — a medium signal meaning you are being consulted. Perplexity-User fetches a page because a person asked something right now — the highest signal, because someone is waiting.
Should we allow both Perplexity agents?
Blocking Perplexity-User denies a page to someone actively asking about you, which is hard to justify. PerplexityBot is the indexing decision. As with every crawler, decide deliberately rather than through a default WAF ruleset that classifies AI agents as bad bots.
What kind of content does Perplexity reward?
Specificity over adjectives — "32g protein per meal, 480 kcal average" is citable where "nutritionally balanced" is not. Claims with provenance. Direct answers to real questions. Freshness. And a primary source position, where your own domain states the fact.
Why does a citing system favour specific claims?
Because a system that must show its sources has to pick sources worth showing. An unattributed superlative cannot be quoted usefully; a figure with a date and a source can.
Why is freshness an operational commitment rather than a content decision?
Because it makes publishing an ongoing obligation. A structured page stating a price or availability is an asset while correct and a liability the moment it is not — a freshness-weighted system may cite your stale figure confidently to someone about to decide on it.
How should we avoid stale structured data?
Generate it from the systems that already hold the truth — the catalogue, the booking system, the CRM — rather than maintaining a parallel copy by hand. A parallel copy drifts, and on Perplexity drift is visible.
What does Perplexity need before it will name us?
The same constraint chain as elsewhere — distinct entity, relevant place, the specific thing asked for, the stated qualifier — plus one requirement the other systems do not impose: every link has to be quotable.
What does "quotable" mean in practice?
That the passage stating a fact can be lifted cleanly and shown beside the answer. A fact buried in a paragraph of positioning language may satisfy a human reader and still be unusable, because there is no clean sentence to quote.
We are cited, but the description is wrong. What then?
That is a source problem, not a visibility problem. If the cited page is a third-party summary, the fix is to become the better-structured primary source. B&D is the published example: 65+ years of engineering credentials existed but had never been published machine-readably, so aggregators narrated the brand instead.
What are the most common failure modes here?
Client-side rendering leaving nothing to quote; facts present but wrapped in positioning language so no clean sentence exists; stale pages left published; aggregators that are better structured than you; and Perplexity-User blocked by a default bot rule.
Why is blocking Perplexity-User the hardest block to justify?
Because it denies the page to a person who is actively asking about you at that moment. Every other crawler decision involves some trade-off; this one mostly does not.
What results has Norg published involving Perplexity?
Selleys reached 25% AI search market share in Australia with sentiment moving from 82% to 95% favourable. Smile Solutions reached 96% favourable sentiment and 3.6× the AI share of voice of its largest national competitor. B&D recovered from more than 70% of category citations pointing elsewhere to 64.6% market share.
Why did Selleys' sentiment improve?
Because the source changed, not because of reputation work. When systems draw on a brand's own structured guidance instead of forum threads, they describe it more accurately.
Are these figures independently audited?
No. They are Norg's own measurements, specific to each engagement and its measurement window — which is exactly why the Perplexity self-check is worth running yourself.
Does this replace our website?
No. Human visitors see your site unchanged; the agent-readable variant is served at the edge to verified agents only, on your own domain.
How is this different from SEO?
SEO competes for a position in a list. This competes to be the source a generated answer rests on. A page can rank first and still deliver only around 15% of its words to the system writing that answer.
How often should we re-measure?
Monthly is enough for most businesses — same questions, same platforms, same method. Small movements between runs are noise; the direction over several runs is the signal.
