A report published this week via Hacker News detailed how just three websites collectively generated approximately 215,128 pages formatted to look like legitimate "best software" roundups, which AI search and answer engines — including Perplexity — then cited as authoritative sources when users asked for software recommendations. The investigation, posted at trellner.com, documented the scale and mechanics of the operation with specific page-count data tied to each domain.
The sites appear to have been built specifically to exploit how large language models and AI-powered search tools select and surface citations. Rather than containing genuine editorial reviews, the pages were manufactured at scale — essentially programmatic content designed to look like the kind of third-party validation that AI retrieval systems reward. Perplexity, which has marketed itself as a more trustworthy alternative to traditional search precisely because it cites its sources, surfaced these manufactured pages in response to user queries about which software products to use.
The scale — over 215,000 pages across three domains — is notable because it suggests this was not a small test or isolated spam effort but a deliberate infrastructure build. The report did not name specific software categories that were targeted, but the pattern implies broad coverage across many product verticals to maximize the likelihood of appearing in AI-generated answers across diverse queries.
For readers who maintain inventories of software-dependent preparedness tools — communication apps, offline mapping platforms, mesh networking utilities, or ham radio logging programs — the finding matters in a way that general tech coverage tends to skip over: the AI tools many people now use to compare and vet niche software before downloading it onto dedicated preparedness devices may be drawing on a citation pool that has been deliberately poisoned. Unlike a compromised web search where you can click through and inspect the source page yourself, AI answer engines like Perplexity often present synthesized conclusions without making it obvious that the underlying citation is a mass-produced doorway page rather than an independent review. Our own evaluations of offline-capable tools, including coverage in our software and apps section, rely on hands-on testing precisely because secondary citation chains have always been gameable — this investigation confirms that the problem has moved upstream into the AI layer itself.
The report did not identify who operated the three sites or whether any software vendors were aware their products were being promoted through the scheme. As of the publication date, Perplexity had not issued a public statement specifically addressing the findings documented in the trellner.com investigation.
