Perplexity

Perplexity is an AI-powered answer engine. You ask a question in natural language, it searches the web, reads a set of sources and writes a synthesised answer with numbered citations linking back to the pages it used. It is one of the main AI search platforms MLforSEO tracks alongside ChatGPT, Gemini, Claude and Google’s AI features.

Not to be confused with perplexity the evaluation metric, which measures how well a language or topic model such as LDA predicts unseen text (lower is better).

How it works

Perplexity runs retrieval for each question, often splitting it into several searches, then uses a large language model to compose the answer from the retrieved passages. Its Pro Search mode shows the sub-queries it ran, which makes it a useful window into query fan-out. Besides the web app, it offers developer APIs and the Comet browser with a built-in assistant.

Perplexity documents two user agents. PerplexityBot surfaces and links websites in its search results and, the company says, is not used to crawl content for training foundation models. Perplexity-User fetches pages when a user’s question needs them.

How it’s used in SEO and marketing

• Visibility tracking. Running a fixed set of prompts and recording whether your brand is mentioned or cited, and where in the answer.
• Fan-out research. Reading the visible sub-queries to see which sub-topics your content needs to answer.
• Crawler access. Checking robots.txt and server logs for PerplexityBot so your pages can be retrieved.
• Analytics. Perplexity passes a consistent referrer, so its visits arrive labelled in GA4, unlike much AI traffic that lands as direct. A custom “AI assistants” channel with a regex such as perplexity\.ai separates it from other referrals.

Example

In the AI Search & Agentic SEO course, Beatrice Gamba suggests writing five to ten queries your audience would ask an assistant, testing each in ChatGPT, Claude, Perplexity and Gemini, and recording whether you were mentioned and whether that was first, middle, last or not at all. Repeated over time, that gives a citation rate you can track.

Limitations and things to watch

• Answers vary between runs, so measure across many prompts and repeat checks.
• Citation does not guarantee a click. A user may read your claim in the answer and never visit.
• Product features, models and APIs change often. Check Perplexity’s documentation for current details.
• Perplexity’s crawler practices have been publicly disputed by some publishers; verify behaviour in your own server logs.

Learn it on MLforSEO

• Blog and Site Architecture for SEO and Conversions in the AI Search Era
• Dark AI traffic: why 70% of your AI influence hides as direct
• 10 Ways to Automate SERP Analysis with AI (monitoring brand mentions in Perplexity answers).
• AI Search Optimisation & Agentic SEO course, lesson on measuring performance in agent-mediated search.