Analysis of AI-Generated Software Recommendations Reveals Source Patterns

A recent analysis across 380 software categories found that 59.8% of sources behind AI-generated recommendations are ranked outside the top 100,000 websites. The study involved querying two AI models for the best products in these categories, resulting in 7,534 citations, with 23.4% pointing to domains not in the top million. Notably, three sites, which collectively published 215,128 machine-generated 'best' pages, were identified as having common operational characteristics. These sites, created after December 2023, have a significant number of citations despite their low web traffic rankings. The median Tranco rank of the cited domains was 71,611, with 36.5% of the cited domains not appearing in the top million. The analysis excluded Google and focused solely on Perplexity's models, emphasizing the retrieval process rather than the quality of the recommendations. The results highlight the prevalence of newer, less-visited domains in AI-generated content and the implications for users seeking reliable software recommendations.