The short answer
Perplexity documents PerplexityBot for search-index discovery and Perplexity-User for user-requested page access, with published IP-range files available for verification workflows.
Business outcome
You can diagnose whether Perplexity access fails at policy, network verification, rendering, or page eligibility instead of guessing.
The process
Build it in five passes
Set policies for both agents
Decide access for PerplexityBot and Perplexity-User separately. Public pages intended for discovery often need both search crawling and user-triggered retrieval to work, while private and authenticated routes should remain protected independently.
Review the wildcard interaction
Check whether a broad `User-agent: *` group blocks paths you meant to expose. Apply the robots matching behavior used by your tester, and verify the final result with real representative URLs rather than visual inspection alone.
Check network controls
Inspect CDN, firewall, bot management, rate limits, geography rules, and origin logs. Perplexity publishes crawler IP information for verification, but operational allowlisting should fetch and validate current files rather than hard-code an old copied range.
Check delivered content
Confirm that a priority URL returns a canonical, indexable 200 response with meaningful text and links. Look for consent walls, empty client shells, soft 404s, blocked assets, or alternate mobile content that changes the page meaning.
Measure downstream visibility
Use a stable buyer-question set to observe recommendations and citations, and keep referral sessions in analytics as a separate signal. Access is a prerequisite to some workflows, not proof that a page will be selected.
Before it ships
Quality checklist
- PerplexityBot and Perplexity-User each have an explicit policy.
- Wildcard rules are tested against priority and restricted paths.
- Current provider IP files are fetched for verification when needed.
- CDN and origin logs distinguish policy blocks from other failures.
- Priority pages return complete, canonical, indexable HTML.
- Citation, recommendation, and referral outcomes remain separate metrics.
Copyable artifact
Perplexity access incident worksheet
Use this when a page is missing or retrieval appears to fail.
AFFECTED URL: [canonical URL] OBSERVED IN: Perplexity search | user request | both FIRST OBSERVED: [date and time zone] ROBOTS CHECK - PerplexityBot decision: allowed | blocked | ambiguous - Perplexity-User decision: allowed | blocked | ambiguous - Matching rule: [exact group and path] DELIVERY CHECK - HTTP status: [code] - Canonical target: [URL] - Indexing directive: [value] - HTML contains primary answer text: yes | no - CDN/WAF action: [allow, challenge, block, unknown] - Origin log evidence: [request ID or none] - Provider IP verification source and fetch date: [details] FIX OWNER: [name] FIX APPLIED: [change] RETEST DATE: [date] OUTCOME: [access restored, still failing, downstream selection unknown]
Validation
How you know it is ready
- 01Both documented agents receive the intended robots decision.
- 02A verified test request reaches the origin or produces an explained CDN action.
- 03The page response contains the same primary facts visible to a human reader.
Do not overclaim
IP verification can support bot-management decisions, but user agents and ranges may change. It does not prove ranking or citation. Always use current provider files.
Questions
What teams usually ask
What is the difference between PerplexityBot and Perplexity-User?
Perplexity documents the first for its search index and the second for user-requested page access. Your policy and monitoring should name both.
Should I permanently copy the published IP ranges?
No. Build a controlled process that retrieves the current provider files, validates the response, and records when the data was refreshed.
Why is a page absent even when both agents are allowed?
Access is only one layer. Indexability, rendering, content quality, relevance, freshness, retrieval systems, and answer selection can all affect the outcome.
Sources reviewed