Member-only story
Why AI Is So Terrible at Handling the News
Any why it’s a big opportunity for creators
News is in my blood. My day job is as the owner of a news photography agency, and my main website is a news site for the Bay Area.
I live and breathe news content. So I’m acutely aware of how bad AI is handling it.
A new study from the Columbia Journalism Review shows that AI search engines like Perplexity and chatbots like Gemini consistently fail to correctly report and cite news stories.
One chatbot that CJR studied — Grok 3 from Elon Musk’s X — got news stories wrong over 90% of the time.
These bots confidently fill in incorrect information, cite syndicated versions of stories on platforms like Yahoo! News instead of the original sources, and even violate publishers’ terms of service by crawling sites where they’ve been explicitly blocked.
The end result? Anyone seeking news content from AI bots is in for a disappointment — or even harmful misinformation.
The Weight
Why are bots so bad at handling news? It comes down to how they’re trained.
Chatbots like Perplexity, ChatGPT, and Gemini are trained on billions of pages of text data, in some cases going back hundreds of years.
