Only what the vendor documents, dated October 8, 2026, then what our weekly checks see Perplexity actually read, and what a business does about it.
Perplexity documents two agents:
Perplexity's help center describes answers as built by searching "in real-time, gathering insights from top-tier sources", compiling "the most relevant insights into a coherent, easy-to-understand answer", with "numbered citations linking to the original sources". It does not describe how sources are ranked beyond "top-tier" and "authoritative".
Sources: docs.perplexity.ai (Perplexity crawlers) and perplexity.ai/help-center ("How does Perplexity work?"), read October 8, 2026.
Perplexity cites more pages per answer than the other three and shows the citations as numbered links, so a brand's presence or absence is visible to the buyer in a way the others hide. Its answers lean on pages with a clear topical authority for the category (the specialist review site, the trade publication, the forum thread where buyers compare), and a brand's own page is cited when it states the fact the answer needs. In the prospects we talk to, Perplexity is the assistant they value least, which is why our free scans lead with ChatGPT and Claude; it stays in the paid weekly check because its citations are the easiest to read and act on.
Rows from scans and checks run in September and October 2026, with the companies unnamed where they have not agreed to be named.
| Category | What happened |
|---|---|
| Weekly check | Perplexity is one of the four assistants in every Core plan: the 50 questions run on it weekly with search on, every answer archived with its numbered sources, and a person grades each flag by the same rules as the other three. |
| Free scans | The free scan runs six to eight questions on two or three assistants, always ChatGPT and usually Claude, because those are the two prospects quote back; Perplexity is added when a prospect's buyers use it. |
| Consulting | On a page check for a consulting firm, Perplexity read the firm's own page when asked directly and confirmed the company was real, then handed the three buying questions to the large firms and a set of boutiques that appear in roundup articles. |
| Case studies | In the four engagements on the results page, Perplexity citations for a telehealth company went from 7 to 3,200 and for a travel publication from 1,400 to 2,400, counted on one independent citation index before and after. |
The 50 questions run on Perplexity weekly, and because its citations are numbered the dashboard shows exactly which pages fed each answer and which of them are yours. The work is the same as for the other three: the page that answers the question, the correction to the record Perplexity quoted, the inclusion request to the specialist site it reads for your category. Where Perplexity cites a forum thread, Grow includes community corrections where the thread's rules allow.
Perplexity says PerplexityBot is the agent that surfaces and links sites in its results and is not used for training; blocking it removes a site from those results. Perplexity-User, the user-initiated fetcher, generally ignores robots.txt by the vendor's own description.
Because a free scan is six to eight questions on two or three assistants, and the two prospects value are ChatGPT and Claude. The paid weekly check always includes Perplexity. If your buyers use it, say so and the scan includes it.
Core corrects pages and records and requests listings; community corrections in forum threads, where a site's rules allow a company to reply, are part of Grow. Either way the thread is traced in the dashboard to the answer it fed.
Or start smaller: the AI Page Check runs one page through all four assistants for $2.99.