AI Search Systems Show Broader Source Diversity Than Google

A comprehensive study conducted by researchers from the University of Duisburg-Essen and the Max Planck Institute for Software Systems has uncovered striking differences between traditional search engines and generative AI search systems in their approach to information sourcing.

Methodology and Scope

The research team conducted 4,600 queries across six major categories including politics, product reviews, and scientific topics. They compared Google's organic search results against four leading AI-powered search systems:

• Google AI Overview
• Gemini 2.5 Flash with search functionality
• GPT-4o-Search
• GPT-4o with search tools

The study provides unprecedented insights into how these fundamentally different technologies process and present online information.

Key Findings on Source Diversity

The most surprising discovery shows AI systems regularly reference websites that don't appear prominently in Google searches:

• 53% of websites cited by AI Overview didn't appear in Google's top 10 results
• 27% weren't even found within Google's top 100 results

This suggests users of AI search tools encounter information from less vetted or obscure sources that wouldn't typically surface through conventional searches.

Source Depth Variations

The study revealed dramatic differences in how deeply systems reference external sources:

• GPT-Tool averaged just 0.4 external sources per answer, relying heavily on internal knowledge bases
• AI Overview and Gemini cited more than eight external sources per query
• Traditional Google searches remain limited to ten links per query, creating more concentrated source pools

Current Events Coverage Gap

The research identified a significant weakness in AI systems regarding timely information:

• When tested on 100 trending topics, AI Overview could only address 3% of queries effectively
• GPT-4o-Search performed better at 72% coverage, but still trailed traditional search engines This limitation raises concerns about users potentially receiving outdated or inaccurate information on developing stories.

Key Points:

🌐 Diverse Sources: Over half (53%) of AI-referenced sites don't appear in Google's top 10 results
📊 Source Depth: GPT-Tool averages just 0.4 external citations per answer vs. 8+ for some competitors
📰 Timeliness Gap: Traditional searches outperform AI systems significantly on current events coverage