Three sites made 215,128 “best software” pages for AI. Perplexity cites them (trellner.com)

Article summary

An experiment with Perplexity's AI models found that 59.8% of the sources used to generate recommendations come from websites outside the top 100,000 most-visited sites. Many of these sources appear to be AI-generated content farms, with three sites producing 215,128 machine-generated pages. The experiment raises concerns about the reliability of AI-generated recommendations and the potential for manipulation. The findings suggest that AI models may be training on low-quality or misleading sources, which could impact the accuracy of their recommendations.

Main themes

• AI-generated content
• reliability of AI recommendations
• manipulation of AI models
• information quality
• AI training data
• content farms
• misinformation

What commenters say

• The rise of AI-generated content is making it difficult to find reliable information online.
• The use of AI-generated sources by AI models creates a risk of exponential amplification of errors and misinformation.
• Some commenters believe that the quality of AI-generated recommendations has decreased over time due to the increasing use of low-quality sources.
• Others argue that AI models can still provide useful recommendations, but require careful evaluation and validation.
• There is a need for better protection against training AI models on AI-generated content to prevent the spread of misinformation.
• The line between legitimate content and AI-generated spam is becoming increasingly blurred, making it harder to distinguish between the two.
• The commercialization of AI-generated content is driving the creation of low-quality sources, which can have negative consequences for users.
• Some users have reported a decline in the quality of AI-generated recommendations over time, citing examples of inaccurate or misleading information.