Perplexity Visibility Audit Checklist: Access, Cited-Source Accuracy and Grounded Tracking

Start the audit by knowing where the vendor is on record and where you are guessing. As of October 2026, this is what the evidence supports.

The gaps shape the rest of the checklist. You can verify access mechanically. You can check accuracy by reading answers. You can't verify why Perplexity chose one source over another, and nobody outside Perplexity can.

Access is the one gate you fully control, which is why it comes first. Our note on fixing crawler access before tracking explains why tracking a blocked site wastes the data.

1. Read robots.txt per bot

Open /robots.txt and find the rules that apply to PerplexityBot. A blanket block on all AI agents is the most common mistake. A minimal allow looks like this:

User-agent: PerplexityBot
Allow: /

Allowing the bot buys eligibility for Perplexity's index. It does not buy retrieval, a citation or a position.

2. Don't rely on robots.txt to block Perplexity-User

Perplexity's own crawler docs say Perplexity-User generally ignores robots.txt rules. If you need to keep a page out of user-triggered fetches, robots.txt is not a reliable control.

3. Test the real response

Request a key URL with the PerplexityBot user agent and record the status code. A 200 passes. A 403, a 429, or an HTML challenge page fails, even if robots.txt says allow. Firewall and CDN bot rules cause this, and the owner usually can't see it from the dashboard. The AI Crawler Checker requests a page as each AI crawler and shows which ones the firewall turns away.

4. Check CDN defaults you didn't set

From 15 September 2026, Cloudflare blocks its Training and Agent crawler categories by default on ad-displaying pages for new domains, new sites on existing accounts and existing free-tier customers, while Search stays allowed by default. Review your zone's AI crawler settings rather than assuming nothing changed.

Vercel's crawler study found AI crawlers fetch JavaScript files but don't execute them, and an independent re-test in June 2026 included PerplexityBot in that group. If your pricing, product names or key claims only appear after JavaScript runs, Perplexity's crawler may see an empty page.

1. Fetch the page with JavaScript disabled.
2. Render it in a browser.
3. Compare the visible text.
4. List every price, number and one-line definition that exists only in the rendered version.

Move those into the server-rendered HTML. A specific fact on a fetchable page gives the engine a sentence it can quote. That is the only on-page change with any positive experimental signal behind it, and it still guarantees nothing.

Perplexity cites whole sources by number, not short text spans. In practice, the pages it cites shape the answer, and many of them won't be yours. If Perplexity is wrong about your company, the cause is often a cited page you don't own.

5. Write down your canonical facts

List the facts that matter: pricing tiers, product names, founding year, headquarters, ownership, integrations. Our guide to brand facts worth giving an AI visibility tool covers what to include.

6. Fix contradictions on your own site

Check that your pricing page, about page and footer state each fact the same way. Contradictions you publish are contradictions an engine can repeat.

7. Audit the cited sources

Ask Perplexity your buyer questions and open every cited URL. Mark each one as accurate, outdated or wrong. When engines cite other sites instead of yours, being named correctly on those sites is the route into the answer. The evidence shows an association, not a promised effect.

8. Watch for manipulated pages

Research on LLM search agents (arXiv 2606.16821) shows that injected instructions on a retrieved page can change what an agent endorses. If a third-party page about your brand contains hidden text or instructions aimed at AI, treat it as a reputational risk. Never use the tactic yourself.

Most Perplexity tracking errors come from measuring an answer that never searched the web. See how API answers differ from consumer app answers for the wider picture.

9. Confirm web search is on

On Perplexity's Agent API, web_search is opt-in. Perplexity's Sonar Chat Completions endpoint retired on 27 September 2026. Code that was not migrated can return answers with no web access and no error. Check every tracking job and confirm it returns search_results.

10. Log retrieved sources, not just citations

The Agent API returns search_results with id, url, title, snippet, date, last_updated and source. Store all of them. The retrieved set tells you more than the cited set alone.

11. Accept the app gap

As of October 2026, secondary reports say the free app uses an auto-router called "Best" and paid plans offer a menu of eight models. No single API model equals the app. Label API results as a proxy.

12. Score Perplexity on its own

In Otterly's 2,000 AI blogs experiment, Perplexity cited the test sites while ChatGPT, Gemini and Google AI Overviews almost never did. A blended cross-engine score would hide that. Report Perplexity separately, the way our ChatGPT visibility checklist treats ChatGPT.

Many Perplexity checklists are content-tweak lists. These items have no evidence behind them as citation levers:

• Schema or FAQ schema. Ahrefs' matched test found no major citation lift on any platform.
• llms.txt. No engine, Perplexity included, documents reading it.
• Chunking or "AI-friendly" writing. The C-SEO Bench study found most rewriting methods ineffective and often harmful.
• Freshness bumps. Changing a date field is self-declared and proves nothing.

Existing markup does no harm. It just isn't a lever. Spend the time on access and accuracy, then repeat the audit each quarter using a quarterly AI visibility review.