Partnerships between news organizations and AI firms could result in a huge loss for journalism
By Jared Schroeder
The Washington Post celebrated a new partnership with ChatGPT parent company OpenAI earlier this week. By doing so, the news organization joined a growing list of firms, including the Associated Press, Wall Street Journal, and Time, that have signed agreements to share content with AI firms.
While these agreements sound like wins for both parties, particularly as cash-strapped news organizations seek new revenue streams, they could end up being tremendous losses for journalism and the flow of information in society. News organizations should think twice, and negotiate carefully, when it comes to these agreements.
There’s no doubt these partnerships are extremely beneficial to AI firms. They are getting high-quality information for their users and troves of data to help train their tools. The benefits for news organizations, and for the future of the flow of information in society, however, are far more fleeting.
Certainly, news organizations are getting a much-needed revenue stream and AI search results are providing links to their websites and content. What could be bad about that? A lot.
In its announcement, the Post celebrated that its partnership with OpenAI will put its content in from of 500 million ChatGPT users weekly. The concern, however, is that a minuscule amount of ChatGPT users will actually click on the links that go back to the Post’s site. Studies have already shown that most people no longer click on search results online because AI summaries and other information that is provided in the result meet their needs.
For news organizations, this means that most AI users will be satisfied with the information they received from the succinct response a tool like ChatGPT provides and won’t click on for more information or to browse other content on the news organization’s website or app.
The deals, in other words, are drawing readers away from the news organization’s actual content. Why subscribe to the Washington Post’s website, app, or print newspaper if a Post-infused AI tool will provide the same information? In this scenario, any subscription money goes to paying the AI firm as users subscribe to its products, rather than the news organization.
News organizations are, in this regard, training their replacements. This is not because AI results are a functional replacement to journalists’ newsgathering efforts. Instead, it is because the news organization is supplying its hard-working journalists’ high-value information to the AI firm and, in doing so, training readers not to go to its news products or that they have to pay for journalism.
Certainly, the AI firms are paying the news organizations for the content in these partnerships. The tradeoff in income streams, from subscriptions and advertising to money from a partnership with an AI firm, probably isn’t worth it.
If news organizations channel their content away from their subscription- and advertising-based models toward AI firms’ products, and users get in the habit of garnering journalistic information from these tools, what happens when the AI firm ends the agreement? News organizations are without a revenue stream and have a weaker economic model because their audiences have been trained to go to the AI tool and not their website, app, or print product. They’ve just channeled the value they provide to their competitor.
We’ve been down a similar road before. News organizations became reliant on Facebook’s popularity with audiences to extend the reach of their reporting. When relationships with news organizations become less profitable, Facebook’s parent company, Meta, moved the social media platform away from news distribution. Traffic to news websites dropped by 33%. Mother Jones, which was receiving about five million website visitors a month from Facebook links, lost 99% of its traffic. That traffic has not returned to news organizations.
Even if these concerns can be resolved, the funding stream from AI firms is not reaching those that need it most—regional and local news organizations. Regional daily newspapers and small-town weeklies have been the hardest hit by changes in journalism’s economic models during the past two decades. More than 3,200 newspapers have closed during that time. About 55 million Americans live in news deserts.
As news audiences receive outsized portions of national news through cable and satellite and from links in social media feeds, local news organizations, which are crucial to holding state and local officials accountable, encouraging democratic participation, and supporting the formation of community, continue to struggle. Partnerships with AI firms, in other words, are not a viable solution to the problems journalism’s business model faces.
Finally, these deals raise concerns about news organizations’ archives. Once a news organization shares its archive with an AI firm, that valuable asset is integrated into the tool’s training and information base. It’s not like a borrowed library book that can be checked out and then given back at the end of the partnership. Thus, news organizations, if they aren’t careful in these agreements, lose an incredibly valuable resource that they could monetize for themselves as they develop their own AI tools.
And that’s likely the best scenario. News organizations should develop AI tools for their organizations, or in partnership with other news organizations, that leverage their precious information for audiences and for their financial models. OpenAI, Google, and other AI firms might be considered information companies, but they are not news organizations. They have a fundamentally different mission. The mission difference is crucial when we consider news organizations’ partnerships with AI firms.
News organizations should be leveraging AI for their audiences and for their business models, rather than being leveraged by the corporations that are developing the most popular AI tools.
———————————————————————————————————————
Jared Schroeder is an associate professor at the University of Missouri School of Journalism and a member of the Overby Center panel of experts. He is the author of “The Structure of Ideas: Mapping a New Theory of Freedom of Expression in the AI Era.”