Google's NotebookLM, an AI-powered research assistant, recently launched a 60-second short-form video generator that converts news articles and other source materials into vertical TikTok-style videos. While the tool promises to make information more digestible, journalists who tested it found that the videos often strip away crucial context, misrepresent sources, and fail to capture the nuance of original reporting.

What Happens When AI Summarizes Investigative Journalism?

When Straight Arrow News tested NotebookLM's video feature on articles from newsrooms across the nation, the results revealed a consistent pattern: the AI excelled at creating engaging, visually interesting content but struggled to preserve the complexity and accuracy of the original work. Kennedy Sessions, a data reporter for The City Reporter, watched as her investigation into New York's treatment of incarcerated people was transformed into a video with unsettling visuals and factual errors.

Sessions noted that the video described one person in her story as 20 years old, when in fact he had been 20 at the beginning of his incarceration, not at the time of the reporting. More troubling to her was what the video left out entirely. "It scared me a little bit," Sessions said. "Because it sounded, I don't know why, but it sounded and it felt different from when you're reading an article and it's like an automated voice reading the article to you." She emphasized that journalism serves an educational role in teaching people about accountability from institutions, something the video failed to convey.

Sessions

"It feels too robotic in a way. Like I think a part of specific investigative work is that human connection and human piece. You can't, like, quantify that in a Google video," Sessions said.

Kennedy Sessions, Data Reporter at The City Reporter

Olivia Marble, a reporter for Spotlight Delaware, experienced similar frustration when her story about the Trump administration ending a federal solar loan program was converted into a video. The original article presented multiple perspectives on the policy change, but the AI video narrowed the narrative to a single viewpoint, excluding counterarguments entirely. "This absolutely did not show the subtleties of what I was trying to convey," Marble explained. She was particularly concerned that the video missed the news peg altogether, failing to mention the Trump administration's role in the loan program's termination.

How Does NotebookLM's Video Tool Actually Work?

NotebookLM functions as a research assistant that summarizes specific pieces of information or writing provided by users. Unlike search engines, it does not browse the web. Instead, users upload or link to source materials, and the AI generates summaries and, now, short-form videos based on that content. The video generator creates 60-second explainers designed to condense important information into formats that appeal to audiences accustomed to consuming content on platforms like TikTok and Instagram.

Jordan Wilson, host of the podcast Everyday AI, noted that this represents a significant shift in how educational AI tools approach content creation. "This is the first time that I've seen an AI product that is focused on education come out with something that is vertical video," Wilson told Straight Arrow News. He suggested the tool could be valuable for businesses creating onboarding videos for new employees, particularly those who learn better through multimedia formats combining text, video, and audio narration.

"A lot of people learn better that way when they can watch something that's interactive or multimedia in terms of text, video and audio narration," Wilson said.

Jordan Wilson, Host of Everyday AI

Where the AI Video Tool Falls Short

The core challenge with NotebookLM's video generator lies in its inability to preserve journalistic precision and complexity. Multiple reporters identified specific problems with how the tool handles source material:

• Factual Errors: Videos contained inaccurate details, such as misrepresenting a person's age or timeline, which could mislead viewers about the substance of the original reporting.
• Missing Context and Nuance: The tool oversimplified complex stories, removing important details and failing to explain the significance of events or policy changes to audiences.
• Exclusion of Key Voices: Multiple perspectives included in original articles were omitted from videos, with the AI focusing on a single narrative thread at the expense of balanced reporting.
• Linguistic Misrepresentation: The AI sometimes used stronger language than journalists had employed, such as describing sources as acting "aggressively" or "desperately" when the original reporting used more measured terms.

Maggie Reynolds, also reporting for Spotlight Delaware, emphasized the importance of precise word choice in journalism. When she watched the video version of her story about a local panhandling debate, she found that the AI had made linguistic choices that exaggerated her sources' experiences. "It's important in reporting to be really precise with word choice," Reynolds noted. "If you're going to use a strong word like that, make sure it's one that is accurate to the details and the evidence in the story".

"I know it's like a 60-second video, so it's shorter than a full length story, but it was a lot of general statements," Reynolds said.

Maggie Reynolds, Reporter at Spotlight Delaware

How to Use NotebookLM's Video Tool Effectively

• Provide Clear Guidance: Users should give the AI specific instructions about which elements of a source are most important, rather than relying on the tool to identify key information automatically without context.
• Review and Edit Output: The generated videos should be treated as drafts requiring human review and editing, particularly when accuracy and nuance are critical to the subject matter.
• Understand Your Audience's Needs: The tool works best for creating quick educational soundbites or onboarding materials where brevity is valued over comprehensive coverage, not for complex investigative or policy reporting.
• Expect a Learning Curve: Wilson emphasized that users need practice to understand which prompts and source materials will yield the best results, noting that "you do need a little bit of practice" to use the tool effectively.

The reporters who tested NotebookLM's video feature acknowledged that the tool has potential value in certain contexts. Sessions recognized that the video "gives people a short summary" of stories, which could be useful for busy audiences. Reynolds noted that she regularly creates one-minute scripts for social media, so she understands the challenge of condensing complex information into brief formats. However, both emphasized that the current version of the tool is not suitable for serious journalism without significant human intervention.

The broader concern raised by these journalists is that AI-generated video summaries might create a false sense of understanding among viewers. "My fear is that someone might watch that and get the false impression that they are understanding every part of what I wrote," Marble said. This risk is particularly acute when the videos are shared on social media platforms where context is easily lost and viewers may not seek out the original reporting.

As NotebookLM's video feature gains adoption, the experiences of these journalists suggest that the tool's success will depend heavily on how it is deployed. For educational content, business training, and quick explainers where nuance is less critical, the tool may prove valuable. But for journalism, policy analysis, and other work where accuracy and multiple perspectives are essential, human judgment and editorial oversight remain irreplaceable.