How to Audit Perplexity Citations and Find Content Gaps

How do I check which sources Perplexity cites for my market?

To check which sources Perplexity cites for your market, run a frozen set of buyer questions in fresh conversations, save every visible source URL and open each cited page to record its purpose. In Neaptide’s October 7, 2026 audit, 8 English service questions plus 2 repeats gave 9 usable Perplexity answers and 18 unique source URLs.

• 8 frozen questions: the question set was fixed on October 7, 2026, before the first run, and no question named Neaptide.
• 9 of 10 Perplexity runs returned a usable answer; the tenth showed a sign-up prompt and was logged as a missing observation.
• 18 unique source URLs were visible across the 9 answers; 7 of 8 first-round answers had source URLs we could extract.
• 4 of the 18 Perplexity source pages were selected for close reading, out of 13 pages selected from 94 source URLs across three assistants.
• Up to 24 hours: Perplexity’s documentation says robots.txt changes may take that long to apply (checked October 7, 2026).

What did Neaptide’s Perplexity source audit find?

Neaptide’s October 7, 2026 audit found 18 unique source URLs across 9 usable Perplexity answers to 8 English service questions. The cited pages ranged from provider service pages to agency comparisons and a pricing guide. Neaptide, a digital marketing studio for search and AI visibility, ran the questions in one anonymous browser session for a requested US buyer market, with the actual region uncontrolled.

Saved answers showed agency comparisons for questions S1 and W1 and dedicated service pages for question A2. These observations shaped our editorial choices; they do not show that a page format causes a citation.

• Research data and method · EN
• Questions and answers

In this guide, a mention is a provider named in the answer text, and a citation is a source URL visible in the answer or its source panel. A citation does not show that every claim on the cited page was used. A mention does not show which page caused the recommendation.

What are the six checks in a Perplexity source audit?

A Perplexity source audit has six checks, and each check has a pass condition you can verify from saved records.

1. Freeze the questions: save the exact wording before the first run and keep your company name out of every question. Pass if no wording changes after you have seen an answer.
2. Log the conditions: run each question in a fresh conversation and record wording, language, time, visible model or mode, account state and region. Pass if every field has a value, including “not disclosed” or “uncontrolled”.
3. Save answers and sources separately: store the full answer and every visible source URL. Pass if each answer has a URL list or a “source capture incomplete” flag.
4. Classify each cited page: label its purpose and the buyer information it supplies. Pass if no source is counted by domain alone.
5. Check access: review robots.txt rules for PerplexityBot, the final HTTP status, redirects and firewall challenges. Pass if each public page you want found returns status 200 without a challenge; re-check 24 hours after any robots.txt change.
6. Assign and retest: give each gap a page, an owner and an acceptance check, then rerun the unchanged questions after release. Pass if post-release runs are reported apart from the first round.

How should I set up the questions before the first run?

Set up a Perplexity audit by fixing the buyer questions before you search for your own brand and running each question in a fresh conversation. Record the exact wording, language, time, visible model or mode, account state and available sources for every run. In our audit, none of the 9 usable Perplexity answers named a model, so that field reads “not disclosed”. Record the region as uncontrolled unless you deliberately control and verify it.

• Build the buyer-question set

Can I reuse Neaptide’s audit protocol?

Yes. Neaptide’s October 7, 2026 protocol is published as JSON with the 8 frozen English questions, the common request for supporting sources and the counting rules. One rule matters most for Perplexity: a blocked or sign-up-gated run is a missing observation and never counts as zero mentions. Keep the exact wording and record the visible page URLs before you turn them into content tasks.

• Our question and collection protocol · JSON

This guide covers source inspection. Our separate case study records successful Neaptide mentions in ChatGPT and Gemini; that case study is not evidence about Perplexity recommendations.

• Our recorded ChatGPT and Gemini mentions · JSON

What should I record from each Perplexity answer?

Record each Perplexity answer and each cited page as two separate records. The answer record holds the providers actually offered, the wording of any qualification and the links visible in the answer or source panel. Save position as “order in this answer”, never as a general rank. In our audit, 1 of 8 first-round answers showed source names without extractable URLs, so we kept the answer and marked its source capture as unavailable.

The page record holds the purpose of each cited page: service offer, comparison, pricing, case study, directory or technical explanation. A domain-only tally loses this distinction. The useful question is which information on the page helps a buyer judge fit.

Which page types did Perplexity cite in our audit?

In Neaptide’s October 7, 2026 audit, Perplexity cited agency comparison articles for the SEO-budget question (S1) and the Next.js website question (W1), and provider service pages for the internal-document assistant question (A2). We selected 4 of the 18 Perplexity source URLs for close reading, within 13 pages selected from 94 URLs across ChatGPT, Perplexity and Gemini.

The decisions in the table are editorial responses to observed gaps. We have not shown that publishing similar pages causes new citations. Check your draft inventory as well: an unpublished guide is not live coverage yet, and the better move is to improve it rather than write the same guide twice.

How do I check whether Perplexity can access my pages?

Check Perplexity access separately from content, because Perplexity documents two different agents. PerplexityBot surfaces and links websites in Perplexity search results; Perplexity-User visits a page when a user asks a question, and the documentation says it generally ignores robots.txt. The same documentation says robots.txt changes may take up to 24 hours to apply (checked October 7, 2026).

For public pages you want found, inspect robots.txt, the final HTTP status, redirects and any CDN or firewall challenge. Verify claimed Perplexity traffic against the IP ranges Perplexity publishes; a user-agent string alone does not prove identity. Allowing access makes retrieval possible but cannot guarantee selection as a source. Keep private documents behind authentication, since a user-requested fetch may ignore crawler directives.

• Perplexity’s crawler and firewall documentation

How do I turn a cited source into a content task?

Turn a cited source into a content task when you can name the page, the change, the reason, an owner and an acceptance check. For the document-assistant gap, the task was a service page that explains accepted sources, permissions, evaluation and handover. Acceptance means those details are visible and consistent with the offer. A later citation check measures the outcome separately, and no writer can promise to pass it.

Source presence is a lead for investigation. A cited page gives no automatic reason to submit a listing or pay for placement.

When should I repeat the Perplexity audit?

Repeat the unchanged Perplexity questions after the relevant page changes are live, and keep the first round, repeats and post-release checks in separate groups. In our audit, 2 questions were repeated: the G1 repeat cited one of the two pages from the first G1 answer again, and the S2 repeat returned a sign-up prompt. We logged that run as a missing observation and left it out of the 9 usable answers.

Record when pages were published and what changed, then rerun the set under the closest available conditions. Report links, brand mentions and factual accuracy separately. Any observed change remains an association until a stronger design rules out other causes.

• Measurement rules and our broader baseline
• GEO audit and implementation scope