The Rise of AI-Generated Journalism

Five years ago, there was no such thing as a newsroom silence. You can see it between the keyboards when you walk through one today, whether you’re in Buenos Aires or London. fewer calls. fewer disputes over a lede.

Staring at a chat window, a junior reporter types a prompt in the same manner that an older colleague once asked a source a question. It’s peculiar because it’s not dramatic.

Tens of thousands of people in six countries participated in the most recent survey conducted by the Reuters Institute, which outlines the shift in numbers that newsrooms have been feigning to ignore. In just one year, the percentage of people using generative AI on a weekly basis increased from 18% to 34%. Now, 22% of people are reached weekly by ChatGPT alone. Journalism has a smaller, almost modest headline figure. Merely 6% of individuals claim to receive news from an AI system on a weekly basis. However, despite its modest appearance, that number has doubled in the last 12 months, and editors believe that doubling is more important than the beginning.

It’s difficult to ignore who is making the request. In order to make a story easier to understand, younger readers—whom publishers have spent the last ten years attempting to entice back with podcasts and newsletters—are now using chatbots. Of those between the ages of 18 and 24, 48% have done this. The percentage is hardly half that for readers over 55. Silently, a generation that grew up scrolling is starting to outsource comprehension. Depending on where you stand on the byline, that may or may not be empowering.

The topic of discussion within the industry has changed from whether or not to employ AI to how to conceal its traces. For the purpose of summarizing court documents or transcribing council meetings, some publications—mostly those with financial resources—have developed quiet internal tools. Others have gone so far as to create whole weather or stock-market wrap-ups without using any human writers at all. A few years ago, there was the Sports Illustrated scandal, the fictitious writers with fictitious headshots, and the peculiar sensation among readers that no one in particular had addressed them. That tale hasn’t really faded. It simply keeps repeating in more subtle, difficult-to-catch ways.

In the meantime, trust is solidifying in a way that should unnerve anyone working in the news industry. Twenty-nine percent of respondents claim to trust ChatGPT. The majority of real news brands would gladly accept that figure. The businesses developing these systems are not journalists, do not use them in any significant proportion, and have no special motivation to defend the reporting profession. However, they are increasingly serving as the public’s conduit for news, summarizing stories, selecting which to highlight, and occasionally making up the rest. Investors appear to think that this is a final result. Editors are less certain.

The smaller items are more difficult to measure. Instead of reading a filing from cover to cover, a court reporter now copies and pastes it into a model. The process by which an audience asks AI to summarize a press release that AI has written. The human component fades somewhere in that loop. Whether anyone is keeping track is still unknown.

Some old-timers like to draw a comparison. Newspapers were supposed to be killed by television. Television was supposed to be destroyed by the internet. Every time, journalism managed to survive by evolving, frequently into something that its earlier practitioners found objectionable. This time, it might or might not follow the same pattern. The factories are operating. The bylines continue to show up. However, the printer is no longer the device in the corner.