Perplexity Cited About Ten Sources per Answer in a September Sample

A September measurement of Perplexity answers puts a number on how many pages an AI answer actually uses. Through 6 September 2026, Answer Radar looked at 67 completed answers. Perplexity searched on every one of them and cited 9.87 sources per answer. The same note puts ChatGPT at 3.35 sources and Claude at 4.04.

Citation is the wide end of the funnel, not the outcome. The write-up splits three events: the page is among the numbered sources, the brand name appears in the answer, and the answer recommends the brand. In 15 customer-project answers, with the publisher's own project left out, Perplexity cited 144 URLs across 98 domains. The customer's domain was cited 7 times. The brand was mentioned in 6 answers and recommended in 2. Six answers recommended nobody.

A citation report can look healthier than the answers

Ninety-eight domains made the panel. Two answers picked the brand. A dashboard that reports citation share would call the presence healthy. A buyer reading the answers would have seen two recommendations out of fifteen. Being cited is the easy step when an engine attaches about ten sources by default.

For GEO, Perplexity's list is long enough that a URL can show up without the brand becoming the choice. Track the recommendation line separately from the footnote line.

Source: Linkeddit, 6 September 2026.