Perplexity

Perplexity runs on large language models rather than a crawled index of ranked pages: its own Sonar model (based on Meta's Llama) and selectable backends such as GPT and Claude search the live web, then assemble one answer with numbered inline citations back to the sources used. Pro Search and Deep Research modes run multiple rounds of searching before answering, and a Search API plus the Comet browser expose the same retrieve-then-cite loop to developers and browsing sessions. It is not a conventional search engine — there is no page-one ranking for a brand to hold — and being cited by Perplexity is not the same claim as being cited by ChatGPT or Gemini, which are general chat assistants where live web citation is not the central product loop the way it is here. Perplexity does not publish how it selects or weights the sources it cites, and its crawlers have separately been accused (by Wired and by Cloudflare) of using undisclosed IPs and spoofed user agents to reach pages that block scraping. A citation inside a Perplexity answer is therefore not a stable position a brand can target the way it targets a search ranking — which is why visibility and GEO tools track it as a distinct, separately-monitored surface.