Answer · feature
How do I track brand mentions in Perplexity?
Last updated · By Perciva Editorial
Short answer
Run a fixed set of buyer-intent prompts in a clean Perplexity session and record three things per answer: whether your brand is mentioned, whether it is recommended, and which URLs are cited. Checking once tells you where you stand; repeating the same prompts on a schedule and diffing the results is what turns it into tracking.
People ask this question in three different ways, and they want three different things. Some want to see, once, whether Perplexity knows them — a five-minute check. Some want to track it continuously, which is a process rather than a check. And some want to know which sources Perplexity is drawing on, because that is where the answer actually comes from. This page covers all three, in that order, because each one builds on the last.
Perplexity is the easiest engine to instrument, and the reason is structural: citation is part of its design. Every answer lists the pages it drew from. That turns monitoring into a concrete loop instead of a guess — if the answer about your category cites a comparison article that leaves you out, that article is a specific outreach target with a name and a URL, not a mystery about what the model "thinks".
The one-off check is simple. Open a clean session — a fresh browser profile or incognito window, signed out — and ask three or four questions the way a buyer would phrase them: "best X for Y", "X vs Y for a 20-person team", "is X worth the price". A signed-in session with your own search history will flatter you, because it has seen you look yourself up. Read the answer and the source list. That is your baseline.
Continuous tracking is a different exercise, and it is where most people stall. The trap is treating it as "check Perplexity occasionally": answers are non-deterministic, so two runs of the same prompt differ slightly, and without a fixed prompt set and a written scoring rule you cannot tell a real change from normal variance. What follows is the procedure that makes the difference readable.
Key facts
- Mention
- Your brand named anywhere in the answer text.
- Recommendation
- Your brand put forward as an answer, not listed in passing.
- Citation
- A source Perplexity lists as having informed the answer.
- Owned citation
- A citation pointing at a domain you control.
How to track Perplexity mentions continuously
Continuous tracking needs three things a one-off check does not: a prompt set that never changes, a scoring rule written down before you start, and a schedule. Everything below runs in a browser and a spreadsheet, and the first pass takes an afternoon.
-
Write 15–25 buyer prompts and then freeze the list
Phrase them the way a person types, with the qualifiers that make an answer specific — team size, budget, integration, industry. Mix category questions, head-to-head comparisons, pricing questions and trust questions. Then stop editing the list. Every number you produce is a ratio measured against this set, so a prompt list that drifts between runs makes the trend meaningless, and that is the single most common reason three months of tracking produces nothing usable.
-
Decide what counts as a mention before you see the first answer
Write the rule down: does "you might also consider Acme" count as a recommendation, or only a named top pick? Does a citation of your changelog count the same as your pricing page? There is no correct answer, only a consistent one — the number is worthless if the definition moves. Deciding in advance also stops you from unconsciously scoring generously on the weeks the results are poor.
-
Run every prompt in a clean session
Fresh profile or incognito, signed out, one prompt per session. Perplexity personalises on history, so a signed-in account that has searched your brand will show your brand. This is the step that silently invalidates most self-run audits, and it costs nothing to get right.
-
Record the answer verbatim, not a summary
Paste the full answer text and every cited URL into the sheet. Summarising as you go destroys the data you will want later: which exact claim was wrong, which sentence named a competitor, which paragraph the citation supported. You cannot reconstruct any of that from a score, and re-running the prompt gives a different answer.
-
Score each row the same day you collect it
Mentioned, recommended, position among the named options, claims made about you, citations and which were yours. Scoring while the answer is in front of you is far faster than a second pass, and it keeps the judgement calls consistent because you are applying one rule in one sitting.
-
Repeat monthly and read the diff, not the level
The useful output is what changed against the previous run on the same prompts: a citation that disappeared, a competitor newly recommended on your top comparison question, a pricing claim that went stale. Monthly suits most categories. Weekly mostly measures the natural variance of a non-deterministic system, which looks like movement and is not.
All of this runs by hand, and the first cycle is worth doing manually so you see what the answers actually say. Automation helps with the repetitive half — running the set and diffing the citations — and Perciva does exactly that; the prompt design and the scoring rule stay yours either way.
The tracking sheet
One row per prompt per run. These are the columns that make the diff readable later — the ones people leave out are the verbatim answer and the citation list, which are precisely the ones you cannot rebuild afterwards.
| Column | What goes in it | Example |
|---|---|---|
| Run date | The date of this cycle, identical for every row in the run. | 2026-09-17 |
| Prompt | The question, verbatim and unchanged from the frozen list. | best password manager for a 20-person team |
| Mentioned | Yes or no — was your brand named anywhere in the answer. | Yes |
| Recommended | Yes or no, under the rule you wrote down in step two. | No — listed under 'also consider' |
| Position | Where you fell among the named options, and how many there were. | 3 of 4 |
| Competitors named | Every rival the answer put forward, in order. | 1Password, Bitwarden |
| Claims about you | Each checkable statement, marked accurate, outdated or false. | "starts at $8/user" — outdated |
| Citations | Every source URL the answer listed. | g2.com/…, acme.com/pricing |
| Owned citations | How many of those were your own domain. | 1 of 6 |
| Answer text | The full answer, pasted. The column everyone skips and later needs. | (verbatim paragraph) |
Two derived numbers are worth computing per run: mention rate (rows with a mention ÷ total rows) and owned citation share (owned citations ÷ all citations). Both are ratios against the frozen prompt set, which is why freezing it matters.
Frequently asked
What is the difference between a mention and a citation in Perplexity?
A mention is your brand appearing in the answer text. A citation is a source Perplexity lists as having informed that answer. They move independently: you can be recommended warmly while every cited source belongs to someone else, which means the description of you is coming from a third party's page rather than your own — and that page is what you would need to change.
How do I see which sources Perplexity used?
They are listed with the answer; Perplexity attaches its sources by design, which is what makes it the most instrumentable engine to monitor. Record the full list, not just whether you appear in it. The domains that recur across your whole prompt set are the ones shaping how buyers hear about your category, and they are a concrete outreach list ranked by influence.
Why do I get a different answer each time I ask the same question?
Because the answers are generated rather than looked up, and the retrieved sources vary. This is normal and it is exactly why a fixed prompt set and a schedule matter: with one run you cannot tell a genuine change from ordinary variance. Running the same set monthly and comparing lets the signal separate itself from the noise.
Do I need to be signed out?
Yes, and this is the step that quietly ruins most self-run checks. Perplexity personalises on your history, so an account that has searched your own brand is far more likely to surface it. Use a fresh profile or an incognito window, signed out, and open one prompt per session.
How often should I track it?
Monthly for most categories, with an extra run after a model update or a significant content release on your side. Weekly tracking mostly surfaces the natural variance of a non-deterministic system, which reads as movement and is not — and it burns the attention you would rather spend acting on what the last run showed.
Can I automate this?
The mechanical half, yes: running a fixed prompt set on a schedule, storing the verbatim answers, and diffing citations between runs is exactly what monitoring tools do. The parts worth keeping manual are the prompt design and the scoring rule, because those are judgement calls about your own market that no tool should make for you.
Is Perplexity worth tracking on its own?
It is the best engine to start with, because its citations make every finding actionable — you can always see which page produced the claim. But visibility differs sharply between engines, so a single-engine picture is a partial one. Start with Perplexity precisely because it shows its working, then widen once the process is running.
See it for yourself
Get a free AI Buyer Perception Snapshot for your brand — what ChatGPT, Perplexity, Gemini, and Claude are saying right now.
Start Free Trial →7-day free trial · No credit card required