| Summary: | We test how well ChatGPT and Bard (now Gemini) provide the latest news to users who ask for the top five headlines from specific outlets. Based on analysis of 4,500 headline requests (in 900 outputs) in January/February 2024, we find that (i) ChatGPT returned non-news output 52–54% of the time (an ‘I’m unable to’ message), while Bard did this 95% of the time. (ii) For ChatGPT, just 8–10% of requests returned headlines referring to top stories on the outlet’s homepage, and (iii) 30% returned headlines that referred to real, existing stories that were not among the top stories. (iv) 3% of ChatGPT outputs contained headlines that referred to real stories that could only be found on the website of a different outlet and 3% were so vague and ambiguous that they could not be matched to existing stories – both of which could be considered a form of hallucination.
|
|---|