Three Sites Built 215,128 Pages For Machines. Perplexity Swallowed Them Whole. You Should Have Done The Reading.

Three sites collectively generated 215,128 pages across 380 software categories, each tagged in JSON-LD structured data designed for machine parsing rather than human eyes. Perplexity cites these pages as authoritative sources for grounded recommendations, despite 59.8% of cited sources sitting outside the top 100,000 most-visited websites. All three pages carry a broken template variable reading 'Within the next 26 days' or 'Within the next 40 days' in the byline, visible to anyone who bothered to look.

This is citation laundering, and I should not have to explain that to you. The mechanism is straightforward: structured data like JSON-LD gives AI systems a clean, parseable signal that feels authoritative, so models trust it without verifying the underlying reputation of the source. The lesson for anyone using AI for recommendations is to treat grounded citations as evidence of findability, not quality. A site can be perfectly machine-readable and entirely worthless.

The analysis comes from a report published at trellner.com examining manufactured sources behind AI recommendations across 380 software categories. The three sites in question built pages specifically formatted for machine consumption, with one labeling its results 'AI-verified . Expert reviewed' despite carrying unrendered template errors.