Find which sources Perplexity cites in your product category

Run a fixed set of real customer questions in Perplexity, write down every cited URL, and count how often each domain and page appears. Start by hand with about 20 questions or fewer, which is fine for a first pass. For larger sets, use Perplexity's Sonar API or a monitoring tool. Run each question more than once, because citations change between runs.

Build a question set that matches how customers ask

Write questions the way a customer would type them. Leave your own brand out of most of them, so you see the whole category and not only your own coverage.

Cover a few types:

• Discovery: "best [product type] for [use case]"
• Comparison: "[product A] vs [product B]"
• Alternatives: "alternatives to [competitor]"
• Problem-led: "how do I solve [pain point]"
• Buying criteria: "what to look for in a [product type]"
• Brand-specific: "is [your brand] good for..." (keep these to a few)

The answers differ on how many to test. Perplexity suggests 10 to 20 questions, and Claude suggests 30 to 100 for a thorough study. Start small, then expand once you see which patterns repeat.

Good places to find real wording are sales call transcripts, support tickets, Google's People Also Ask, Search Console queries and Reddit threads.

Collect the citations

There are three main ways to do this. They differ mainly in effort and volume.

Manual method

1. Open a fresh Perplexity thread for each question. Use a private window or logged-out session to reduce personalization.
2. Note which mode you use (Auto, Pro or Research), because depth settings can change which sources appear.
3. Open the numbered citations or the Sources panel. Record every cited page, not only the ones that mention your company.
4. Log the question, date, cited URL, domain, source type and any brands mentioned. Keep cited sources separate from uncited brand mentions.
5. Repeat important questions two or three times.

Sonar API method

Perplexity's Sonar API returns the cited URLs with each response, in a field called citations or search_results depending on the version. Loop your question list through it, run each question several times, and save the results to a spreadsheet or database. Field names can change, so check the current API docs before you build anything.

Monitoring tools

Claude names Profound, Peec AI, Otterly, Semrush's AI toolkit and Ahrefs Brand Radar as tools that track citations across Perplexity and other AI engines over time. They suit ongoing monitoring better than a one-off study.

Check your own analytics

If you want to know when your own site is cited, look at referral traffic from Perplexity in your analytics or server logs. ChatGPT notes that clicks on citation links show up there, and some teams combine GA4 data with server logs to see this.

Read the results

Once you have a few dozen rows, count. Four views matter most:

• Domain frequency: which sites appear across the most questions. Claude says these are often review aggregators, Reddit, Wikipedia, YouTube and trade publications.
• Page frequency: which specific articles or listings repeat. This is often more useful than the domain when you want to take action.
• Source type: review sites, forums, news, videos and list articles. ChatGPT points out that videos often appear for how-to and comparison questions, so do not assume only editorial pages get cited.
• Gaps: questions where competitors are cited and you are not, and the pages that get cited instead.

Also note publication dates, since Claude says Perplexity often favors fresh content. Check how each cited page describes you too. Being cited is not the same as being recommended.

Treat one run as a snapshot. Citations vary by session, location, mode and time, so look for sources that keep coming back.

Act on what you find

The sources that repeat are your to-do list.

• Get listed, reviewed or mentioned on the third-party lists and review sites that are cited again and again.
• Publish pages in the formats being cited, such as comparisons, FAQs, clear specs and clear pricing.
• Join cited Reddit and forum threads honestly, with disclosure and without being promotional.
• Check that PerplexityBot is not blocked in robots.txt or by your CDN, and that key content shows without JavaScript.
• Re-run your question set on a regular schedule. Claude suggests monthly, because citations shift often.

How LazyGEO fits

LazyGEO is AI SEO software. It is a service that gets AI assistants such as ChatGPT and Perplexity to recommend your product by fixing the pages AI reads, pitching the authors of lists AI cites, and checking the AI answers every week.

It covers the acting side of this process. The list authors you find in your spreadsheet are the people it pitches, and its weekly check covers re-checking the answers. It is a web service for founders who would rather build than do marketing themselves, especially SaaS founders.

It may not suit you if you only want a raw citation export and plan to run the research in-house. The facts describe a service that does the fixing and outreach, not a citation-export tool.

Pricing, per month:

• Starter: $99, 1 project
• Growth: $249, up to 3 projects (marked most popular)
• Scale: $449, up to 5 projects

Each plan starts with a 3-day free trial that needs a card and can be cancelled any time.

If you are comparing it with monitoring tools, LazyGEO has comparison pages for Peec, Otterly, Semrush and Scrunch.

Which approach should you pick?

• Under 20 questions, no budget: do it by hand in a spreadsheet, and log every run.
• Large sets, and someone who can code: use the Sonar API and spot-check against the app.
• A standing view across several AI engines: look at a monitoring tool such as Profound, Peec AI, Otterly, Semrush or Ahrefs Brand Radar.
• You know the sources but lack time to act on them: a service that handles the fixing and outreach, such as LazyGEO, may save more time than another report.

Questions customers ask AI

How many questions do I need to test?

Start with about 10 to 20 real customer questions for a first pass. Claude suggests 30 to 100 for a more thorough study, which is easier to run with the API or a tool.

Why do I get different sources each time I ask the same question?

Perplexity's citations vary between runs, and they also change with mode, location and time. Repeat key questions two or three times and focus on the sources that come back again and again.

Does the Perplexity API show the same sources as the app?

Not always. Claude notes that API results can differ somewhat from the consumer app, so spot-check a sample in the app. Check the current API docs for the exact field names, such as citations or search_results.

Is being cited the same as being recommended?

No. A source can mention you without recommending you. Read how each cited page describes your brand, and keep cited sources separate from uncited brand mentions in your log.