how can i monitor perplexity brand mentions

A brand mention is an identifiable reference to your business or product in the answer text. A citation is a source link. A recommendation is an answer that puts your product forward for the buyer’s task. These can occur together, but they are different observations. Decide whether subsidiaries, former names, and product names belong in your count.

For a generic brand name, require context such as the product category or domain. A common dictionary word appearing in an unrelated paragraph is not a reliable match. Keep uncertain matches in a separate review column rather than counting them as wins. This makes later comparisons much easier to defend.

1. Start with real tasks. Take questions from sales conversations, support requests, and your own search data. Write them in the language a prospective customer would use.
2. Separate discovery from branded questions. “Which scheduling tools work for a small clinic?” tests discovery. “Does this product support clinic scheduling?” tests knowledge of a named product.
3. Add decision constraints. Country, budget, integration, and team size can change which alternatives are relevant. Record those constraints in the prompt itself.
4. Freeze a manageable set. Ten to twenty well chosen questions can support a useful starting log. This is a workflow suggestion, not a statistically representative sample of all Perplexity users.
5. Record changes. If you add a question later, put it in a new cohort. Otherwise an apparent improvement may simply reflect an easier question mix.

Use a new conversation for each question. Record the exact prompt, the date and time, the selected mode or model when visible, language, and any location context. Save the full answer and the cited source URLs. A screenshot is useful for checking what appeared on screen; a text export makes counting easier. Respect the platform’s terms and access controls.

For each answer, mark mention yes or no, owned-domain citation yes or no, recommendation yes or no, and the competitors actually named. Record which URL supports each claim. Do not convert every external citation into a “cited competitor”: a review site, a public agency, or a documentation page may be a source without being a competing product.

Check ambiguous cases manually. For example, a citation might point to a page that compares several vendors while the answer recommends only one. The citation alone does not prove that all vendors were recommended. Preserve the distinction in your log.

Choose a cadence that fits the decision you need to make, such as a weekly review while updating a set of product pages. Keep prompts, language, and product surface stable. Report completed answers and failed requests separately. A timeout is missing data; it is not evidence that the brand was absent.

In a hypothetical log of 20 completed answers, eight answers mention the brand and three link to its website. The mention rate is 8/20, or 40%, and the owned-site citation rate is 3/20, or 15%. These figures describe that log only. They do not estimate your share of every answer shown to every user.

Compare the original question cohort with itself. If the model or search mode changes, mark the break in the record. A single changed answer is a reason to inspect the underlying sources, not sufficient proof that your last website edit caused the change.

Before buying software, ask to inspect a raw answer export and the exact product surface it samples. Our tool comparison offers a selection checklist. This site’s live API check is currently unavailable; no sample answer shown here is presented as a live test.

If your pages are never cited, first check whether the relevant public page is accessible and explains the buyer’s task clearly. Perplexity’s official crawler guide identifies PerplexityBot for search discovery and distinguishes it from the user-triggered Perplexity-User fetcher. Follow those documented controls when reviewing access; publishing an llms.txt file is not evidence that the search product will read it.

Then inspect the sources behind competitor recommendations. Look for a concrete information gap you can address accurately, such as an undocumented integration or an unclear eligibility requirement. Keep a record of the change and recheck the frozen questions. Measure actual visits separately using GA4 referral reporting.