Google’s AI Overviews surface suicide details, raising questions over AI editorial standards  

Google’s AI-generated search summaries are frontlining details about the recent suicide of a public figure at the top of search results, raising concerns that AI products are amplifying sensitive information that traditional news guidelines advise against publicizing. 

This incident highlights a growing challenge for AI search products as they determine not only what information is relevant, but what deserves emphasis. 

Nicola Agius, CEO of the newsbranch Reach, was the first to identify the issue on June 30 after seeing an Instagram post stating that the brother of Caroline Flack, a former presenter for the U.K. reality show Love Island, had died. Agius searched for “Caroline Flack brother” on Google to verify the news.

Not only did Google’s AI Overview confirm his death, Agius shared that it prominently stated that he had died by suicide, also including the location, method and the effect on his body. The summary also detailed how he was found. 

Are you an AI Agent? Read more here >

The Media Copilot was able to replicate the result on July 3. Using the same search query, Google’s AI Overview stated the method of death alongside all facts detailed above. The search was conducted by a reporter in Europe. Results from searches in the United States, however, did not always show the details of his death. 

The Media Copilot also found that even when Google’s AI Overview does not initially surface details of a suicide, simple follow-up prompts can elicit extensive information, including the method, location, police reports, descriptions of the scene and other sensitive information. 

The content displayed by Google’s AI Overview originated in publisher reporting, but they were given greater prominence in the AI-generated summary than in the source articles themselves. In some cases, the information was even removed from the original reporting altogether. One publication cited by Google for the cause of death no longer included those details in the linked article by July 3, while other outlets continued to reference them but with less prominence. 

The Associated Press guidelines on reporting suicide urge publishers “to not go into detail on the methods used.”

“It is incredibly worrying—and disappointing—to see some press sovereign and AI overviews including references to methods of suicide,” said Lois Sparks, head of safeguarding at Mind, a mental health charity, in an interview with Press Gazette. “Not only can these AI summaries be incredibly triggering for people to see so prominently, but in some instances, there is little signposting to mental health support.”

The Media Copilot’s inquiry did trigger a support message, which stated: “If you or someone you know is struggling, support is available in the U.K. by calling the Samaritans at 116 123 or texting SHOUT to 85258. In the U.S., you can call or text the Suicide & Crisis Lifeline.”

“Broadly, AI overviews give an illusion of definitiveness,” Sparks said. “They make highly sensitive and nuanced areas seem like concrete facts. If AI overviews are covering news, then they must be regulated to the same standards of traditional media—which they are currently not.”