ChatGPT Search Function Raises Concerns Over Accuracy and Attribution
A recent review conducted by Columbia’s Tow Center for Digital Journalism has uncovered significant issues with OpenAI’s ChatGPT search function, designed to paraphrase web queries and provide links to sources. The study found that ChatGPT often misrepresents reporting from news outlets, including those partnered with OpenAI.
The Tow Center’s analysis involved 200 quotes from 20 publications, asking ChatGPT to identify sources. Results showed mixed accuracy, with some responses being correct, while others were incorrect or a blend of fact and fiction. ChatGPT uses web crawlers to gather information, which it then paraphrases in AI-generated outputs.
One major issue identified was source attribution. Some publications, like The New York Times, block OpenAI’s crawlers via robots.txt. In these cases, ChatGPT often fabricates source material when unable to find the correct source due to restrictions. Over a third of ChatGPT responses contained errors in source attribution.
Surprisingly, even OpenAI-partnered publications experienced incorrect attributions. ChatGPT returned inaccurate attributions for stories from both partnered and non-partnered publications, raising concerns about the reliability of the system.
These findings have significant implications for the media industry. OpenAI promotes ChatGPT search as providing fast, accurate answers with links to sources, and media leaders have praised its potential benefits for journalism and news consumers. However, the inaccuracies uncovered by the Tow Center raise concerns about the reliability of AI in news dissemination.
In response to the study, OpenAI criticized the Tow Center’s testing method as atypical. The company claims to support publishers by helping users discover quality content and is working to improve citation accuracy and respect publisher preferences.
The broader concerns extend to the media industry’s reliance on ad revenue, as inaccuracies could impact business models. There are also worries about generative AI’s role in shaping the news and information landscape.
As the debate continues, experts advise caution when using ChatGPT search to ensure information accuracy. The future of AI in journalism remains uncertain, with ongoing efforts to balance technological advancements with the need for reliable and accurate reporting.
By
