Synthetic Sources?: Auditing Generative Search Engine Citations for Evidence of AI-Generated Sources

Abstract

The growing accessibility of Large Language Models via conversational interfaces capable of responding to users’ questions by drawing on, synthesizing, and citing information from the web (i.e., Generative Search Engines) has simpli- fied the information-seeking process for users. However, with the proliferation of AI-generated content on the web, it is un- clear whether these engines can reliably omit citing synthetic sources (i.e., AI-generated sources). Should these engines be unable to do so, this puts users at risk of harm by treat- ing information from AI-generated sources synthesized in re- sponses of generative search engines as equivalent to infor- mation from authoritative or official sources. In a step towards identifying whether AI-generated sources are being cited by these engines, this work presents an audit of four generative search engines (ChatGPT, Copilot, Gemini, Perplexity) us- ing a total of 712 real-world human-generated queries span- ning domains of public importance: politics, health, and the environment. Our findings show evidence of AI-generated sources being cited across all four generative search engines (∼16% of cited sources) and identifies key source web do- mains these sources belong to that are frequently cited across these engines and topics. In addition, we observed that gen- erative search engines include a somewhat narrow set of re- peatedly cited domains while predominantly surfacing a large number of minimally cited domains in responses to users’ queries. These findings contribute to the growing body of work on assessing the risks of generative search engines with the objective of increasing public awareness of their limita- tions and encouraging appropriate measures to improve in- formation quality and governance of these systems.