The rise of AI-generated content may lead more people to seek out validated news sources, research from the National Bureau of Economic Research (NBER) reveals.
Traffic to the news site rose 2.5 percent after readers were given a quiz to identify AI-generated images, highlighting trust garnered from audiences when educated about AI content.
The research, titled GenAI Misinformation, Trust, and News Consumption: Evidence from a Field Experiment, collected data from an experiment conducted by German news outlet Süddeutsche Zeitung (SZ).
The experiment presented SZ readers with a series of images. Readers were asked to identify which were AI-generated and which were not. They were then questioned about their trust in media amid the AI age.
Authors Filipe R. Campante, Ruben Durante, and Felix Hagemeister, and Ananya Sen shared that only 2 percent of quiz-takers correctly identified which images were created by AI. Over one-third (36 percent) answered none of the questions correctly.
The readers then largely expressed concerns about misinformation and a decreased trust in news sites.
However, this activity of educating readers drove more trust within SZ readers.
Following reader participation in this quiz, the researchers observed the 2.5 percent increase in visits to SZ’s website. Subscriber retention also increased 1.1 percent over five months, the research revealed–about a one-third decrease in subscriber loss rate.
The researchers theorise that, with an increased prevalence of AI content, news consumers will return to more traditional consumption methods.
“In an environment with widespread misinformation, a sufficiently trustworthy news source – in the sense that the consumer expects it to do a sufficiently good job mitigating the challenge posed by misinformation – becomes relatively more valuable…” the paper read.
“That is because what matters to the consumer’s choice is the source’s value relative to the alternatives, which are degrading more rapidly.”
A Sea of Deepfakes
The advancement of AI content generation tools increases the potential for inaccurate or fake “news” content that could be mistaken as legitimate by the public. Tools like Google’s Veo 3 allow users to generate film-quality videos with synchonised audio.
In June, a month after Veo 3’s release, online verification platform GeoConfirmed reported an increase in AI-generated misinformation. This included fake videos of air strikes in Israel and Iran.
The likelihood of AI-generated fake news increases as more people rely on social media alone to get news. More than half of people (54 percent) in the US get their news from platforms like Facebook, X, and YouTube, the Reuters Institute reported in June. Without abiding by the standards of formal news outlets, users on social media platforms can easily spread fabricated content and pose it as real life events.
However, Google has been making some efforts to counter the misuse of their AI products. The company released SynthID, a watermarking tool aimed at identifying AI-generated content.
On Veo’s site, under a tab labeled “Safety and Responsibility”, Google states that the company “blocks harmful requests and results”, and that “Veo outputs will undergo safety evaluations and checks for memorised content to reduce potential issues related to privacy, copyright infringement, and bias.”
