In short. Perplexity is an AI answer engine that searches web sources and returns conversational answers with numbered source links.

Definition: Perplexity is an AI answer engine that searches the web and produces conversational answers with links to source pages. Its help center describes responses that combine information from multiple sources.

Key facts

As of 17 October 2026, Perplexity's search documentation describes numbered citations that link to original sources.
Perplexity's help center says follow-up questions retain the context of previous queries.
The crawler documentation identifies PerplexityBot as the agent used to surface and link websites in search results.
The same documentation identifies Perplexity-User as an agent that may fetch a page in response to a user's question.

How does Perplexity produce an answer?

Perplexity's explanation of its search process describes interpreting the question, searching online sources and summarizing the information into an answer. Numbered citations connect the response to source pages, giving readers a way to inspect the material behind it.

The same page distinguishes Pro Search, which supports model selection, from Research, which performs repeated searches and reads sources to produce a report. It also describes follow-up questions that preserve conversational context. An answer therefore belongs to a particular question and conversation; when reviewing a response, recording the follow-up questions helps preserve that context.

Which agents access website content?

Perplexity's crawler documentation describes separate roles for automated search access and user-requested page retrieval. According to Perplexity, PerplexityBot supports search results and does not crawl content for training foundation models, the underlying AI models used by applications.

The documentation recommends allowing this agent in robots.txt, the file used to communicate crawler access rules. An example implementing that recommendation is:

User-agent: PerplexityBot
Allow: /

Perplexity-User may visit a page while answering a question and include its link in the response. Perplexity says this fetcher generally ignores robots.txt because a user requests the fetch, and that it does not collect content for foundation-model training.

How can a website check access and source links?

Perplexity's access guidance recommends checking both the user-agent identifier and published IP ranges when configuring a web application firewall, a system that filters incoming website requests. It publishes separate JSON address lists for PerplexityBot and Perplexity-User and recommends monitoring server logs after changing firewall rules.

For reviewing an answer, the concrete evidence is its linked source page. Perplexity's citation guidance describes opening those links to verify information. A practical review records the question and opens each relevant citation to compare the answer's wording with the page. Server logs document access requests; the answer itself shows which links appear.

FAQ

How can you inspect an answer's sources?

Open the numbered citations. Perplexity's help center describes these as links to original sources, which you can read to check the information in the answer.

Does blocking PerplexityBot also control user-requested fetches?

Perplexity's documentation describes Perplexity-User separately and says it generally ignores robots.txt. If you manage access through a firewall, its guidance covers separate agent identifiers and IP address lists for these requests.

Sources

Check how a page reads to answer engines. Open the GEO score analyzer