The Jubillar case — involving the disappearance of a French woman and her husband's eventual murder conviction — is one of the best-known true crime stories in France. With the story unfolding near Toulouse, the region’s newspaper of record, La Dépêche du Midi, became an expert in the case. “We had a lot of journalistic knowledge and information in our archives, photos, and videos that other media in France didn’t have,” explains Christophe Rauzy, La Dépêche du Midi’s deputy editor-in-chief.

With the trial beginning in 2025, Rauzy and his colleagues knew the entire country would soon be focused on their region.

But La Dépêche du Midi’s greatest strength — its deep archive — was also its biggest challenge. Years of reporting were siloed across 900 articles, 500 pages of confidential documents, and over 1,000 internal notes. “We had the information,” explains Christophe Rauzy, Deputy Editor-in-Chief, “but we couldn't see the hidden connections. We needed a way to take a step back and see the whole picture.”

Google introduced La Dépêche to NotebookLM. A team of journalists began using the tool to help interrogate their collective knowledge. Unlike many AI tools, NotebookLM operates in a closed environment, drawing only from documents uploaded by the user, which are not used to train the underlying models. “That was a central question for all of my colleagues,” Rauzy reveals. “We needed to ensure that the documents weren’t accessible to anyone and that our sources were well-protected.”

With their security questions answered, La Dépêche’s staff participated in NotebookLM training from the Google News Initiative (GNI). They quickly discovered that the tool could help them reconstruct complex chronologies, reveal information buried in disparate documents, and suggest angles based on automated mind maps. For instance, they pulled a thread that led to a forgotten detail — the licence plate of a camper van that allowed them to track down a witness and produce a completely new article on a cold lead during the investigation.

To prepare for a key interview, the team usedNotebookLM to structure all the evidence against the suspect from their archive. This produced a balanced, argument-based summary that helped them formulate tougher, more precise questions. “It allowed us to argue with a better equilibrium of the facts,” Rauzy notes.

NotebookLM also served as an instant knowledge-sharing base during the trial, allowing journalists to immediately pull validated, archived information to aid in their live reporting. The tool could synthesize information across hundreds of sources in seconds. This helped the team to create in-depth explainers, like one tracing the entire history of a single piece of evidence, that would have been too time-consuming to produce manually.

By synthetizing information across hundreds of sources in seconds, Rauzy estimates that NotebookLM saved his team two hours per day on the Jubillar case, and their edge in live reporting resulted in a number one search ranking among French media at the time of the trial, from September through October 2025.

In September 2025, La Dépêche and GNI presented their use case of NotebookLM at the Festival de l’info locale, an annual news festival dedicated to local media. “I was overwhelmed with questions from other journalists at the festival about using NotebookLM,” Rauzy says, “and many of them wrote to me afterwards because they wanted to know how to use it.”

The true prize, however, is the introduction of a new newsroom tool. “Now we know we can use it for many things,” Rauzy says. La Dépêche has begun incorporating NotebookLM into their everyday operations, using the tool to verify ideas at their editorial meetings and to gather granular data. Rauzy also points to sports reporters who can now "have a conversation with their binders," instantly cross-referencing player data to find fresh angles. "It’s a new tool at a journalist's disposal, one that supports our thinking."