Study: Perplexity Grounds 'Best Software' Answers in 215,128 SEO Pages

An investigation that queried Perplexity's sonar and sonar-pro models across 380 buyer-intent software categories found that 59.8% of the 7,534 citations returned point to domains ranked worse than #100,000 on Tranco, and 23.4% aren't in the top million at all; the median citation rank is 71,611. This isn't a story of a few dominant sites — 36.5% of the 2,055 cited domains never crack the top million, and they skew far newer than the well-ranked ones.

At the center appears to be a single coordinated operation: wifitalents.com, worldmetrics.org, gitnux.org and zipdo.co, all registered after December 2023, all sharing the same Cloudflare nameserver pair and page template, and all titling their homepages "Facts & Grounding Page" — grounding being the retrieval step these models perform before generating an answer. Together the four sites have published 215,128 machine-generated "best <category> software" pages, far exceeding any real number of categories. Fetching the same category page across three of the sites produces contradictory top-five rankings, three different named "editors" per site, and an unrendered template placeholder left in the byline on every page.

Separately, a demo-software vendor's marketing blog — with no stake in any category it was cited for — became the third most-cited source overall, ahead of Gartner, purely on volume. Of 1,502 vendor homepages checked, 17 were dead or unreachable.

The takeaway for engineers: RAG-style search layers can be systematically targeted by content farms built explicitly for machine consumption rather than human readers. Citation quality and source verification need to be treated as a first-class part of any production grounding pipeline, not an afterthought.