daBongo LMS AI Training Courses

Lesson Content

How Perplexity generates a response.

When you submit a question, Perplexity does three things in sequence:

Stage 1 – Search: Perplexity queries the web to find relevant sources. The sources it finds depend on which sources are indexed, which ones are accessible, and how the search query is constructed from your question.

Stage 2 – Reading: Perplexity processes the content of the retrieved sources. It reads what is publicly available – it cannot access paywalled content, private databases, or sources that block automated access.

Stage 3 – Synthesis: An AI model synthesizes what it read from the sources into a coherent response. The AI writes a summary in natural language, attaches citations to claims, and presents the result.

Where errors enter at each stage.

Stage 1 errors: Perplexity may not find the best or most authoritative sources. It can miss important sources. The sources it finds may not be representative of the full picture on a topic.

Stage 2 errors: Perplexity may not have full access to source content (paywalls, robots.txt restrictions). It may read recent content that is itself incomplete or preliminary.

Stage 3 errors: The AI synthesis may misrepresent what sources say. It may conflate information from different sources. It may overstate certainty. It may miss nuance that is clear in the original source.

The critical implication.

A Perplexity citation means: "My synthesis of this claim was influenced by content from this source." It does not mean the claim accurately represents the source, that the source is authoritative, or that the claim is true.

This is why reading the citations – not just trusting the synthesized summary – is a non-negotiable professional habit.