Generative AI citation and source retrieval patterns

Research into generative AI systems indicates significant fragmentation and inconsistency in how different platforms select and cite digital sources. Studies show that citation behavior varies based on the specific AI engine, the topic queried, and the accessibility of the source material, with systems often favoring public pages with stable URLs over gated or complex files.

In the cryptocurrency sector, the citeOS Exchange Index identified a disconnect between market volume and AI recommendations. For example, while Binance maintains a daily spot volume approximately five times larger than Coinbase, Kraken was the most frequently recommended exchange across ChatGPT, Perplexity, and Gemini. Rankings remained inconsistent, as an exchange might appear on one chatbot’s shortlist but be absent from others.

Further analysis of the Austrian tourism sector revealed a highly fragmented digital source market. A study evaluating 1.14 million references found that no single platform dominates, with the most frequent provider accounting for only 3.7 percent of references in August 2026. In accommodation queries, booking portals provided 37.6 percent of citations, while direct business websites provided 28.7 percent. The research also noted technical variations between systems, such as Google-based applications relying more heavily on social media content compared to other models.