Introduction
The distribution of information is experiencing a seismic shift with the integration of artificial intelligence, fundamentally transforming how news is produced and consumed. A recent partnership between OpenAI and News Corp signifies a significant development in the intersection of these generative AI systems and journalism, allowing the industry to move faster and re-invent its age-old processes through enhanced content creation and automation of writing or ideation tasks.
The potential benefits of this partnership (and those similar) are substantial: AI can streamline news production, generate articles, summaries and headlines with remarkable efficiency, and reduce operational costs for media organizations. However, this technological integration comes with the risks of intensifying the existing issues with bias, misinformation and ownership in generative AI, as well as reforming how consumers interact with information. The reform has so far been lauded as a way to make information distribution more accessible and efficient, but long term impacts of streamlining news processes with AI will negatively affect public access to knowledge.
Current State of Relying on AI for News
In May of 2024, OpenAI entered into a significant multi-year partnership with News Corp, allowing OpenAI's ChatGPT to utilize content from News Corp's vast array of publications. This agreement grants ChatGPT access to both current and archived content from notable News Corp outlets such as The Wall Street Journal, Barron's, and The New York Post in the U.S. The partnership is valued at approximately $350 million, aiming to supplement ChatGPT' output with high-quality, reliable journalism.. This partnership is also part of OpenAI's broader strategy to compete more effectively with other tech giants like Google by becoming a primary gateway for information and news. For media organizations like News Corp, the deal presents an opportunity to utilize AI to enhance news production processes, engage audiences more effectively, and generate new revenue streams. However, despite these efforts, many challenges persist through the partnership including bias, public trust and credibility, and hallucination errors.
In a general sense content creation is immediately more efficient by utilizing the tools and efficiencies made possible by large language systems. Already, publications like the Associated Press and the NYT have made motions to dedicate teams for determining the best ways to use AI to supplement their practices, through both the ideation and writing stages to reported great success. In doing so, despite their reassurances that their human journalists will not be replaced, they take a step towards greater efficiency; a cheaper, faster, and arguably equally effective manner of content generation. Furthermore, the institution of intelligent algorithms has already long been touted as a competitive advantage for these sites, as they allow sites to curate specific content for their users, maximizing viewership and engagement for the site, while increasing their customer satisfaction. As technology improves, the dependance and utilization of these AI systems will also become more significant and widespread. As media organizations navigate this landscape, they must strike a balance between using AI's efficiencies and maintaining the critical role of human judgment in journalism.
While the use of AI as a tool is not problematic itself, dependence on algorithms in every stage of the information process may have wider impacts. On the content generation side, current generative systems continue to be victim to hallucination errors, presenting information as factual (even going as far as sourcing it), when the data simply does not exist. While additional training data alleviates the risk of hallucination, current state systems are unable to bring the risk low enough to be able to defend AI generated content being distributed as a tool for public information. As long as that risk is prevalent in AI tools, it poses two distinct issues: greater misinformation proliferating through society, as well as a serious threat to the integrity of news reporting and public trust in media, especially when consumers are unable to distinguish between AI-generated content and human-authored journalism (by a recent study conducted by Nexcess concluding that participants (aged 18-24) were only 61% successful in distinguishing AI content, older age groups being even less accurate). The outputs of generative systems should also be critically examined: AI systems rely heavily on the data used to train them, and if this data is incomplete, biased, or unrepresentative, the result will be skewed or inaccurate. This bias can perpetuate existing prejudices or distort coverage of sensitive topics, undermining the goal of objective and impartial news reporting - an issue exacerbated if readers assume that data-trained systems are unbiased by nature. As AI tools are capitalized on for distribution purposes as well as content generation, their incorporation contributes to an increasing echo chamber. Social sites and media organizations already take advantage of complex algorithms and user data to deliver to their customers the content that is most engaging to them based on historical preference and behaviour. Movement toward trend-following behaviour driven by AI engines may lead to a homogenous set of content as stories that are more attention-drawing are prioritized, rather than those that are less popular but more impactful. The resulting content bubble limits consumers access to diverse sources, and visibility of content changes based on the data collected from users. As intelligent automation becomes more integrated in every point in the content lifecycle, echo chamber potential only grows and becomes more significant: with engagement motivated algorithms driving the creation process, engagement-driven content is the result, rather than journalistic integrity and investigatory focused pieces.
Future State
As such, these risks and benefits stem from current state technologies and applications, without consideration for the further steps being taken to advance artificial intelligence systems and incorporate them even further into the lives of consumers. In many cases, this integration magnifies the effects of the risks and harms of AI without similar emphasis of the benefits that it provides. These negative implications have the most substantial impacts in the areas of active spread of misinformation, and the limiting of opportunity for actual journalist-led news organizations.
As corporations focus on delivering information directly to consumers, from question to answer, the risk of misinformation grows. Different AI systems, such as Perplexity - a chatbot-powered search engine - aim to provide quick answers, bypassing the need for users to visit the original sources. This approach can exacerbate the problem of misinformation, as users receive information stripped of context and background. The risk of hallucination is crucial in this context. While AI systems might draw on larger data pools, they can still pull incorrect information even if it is not outright hallucinated. For instance, according to a study published in Electronic Markets, Natural Language Processing AI models assign different weights to different sources, which significantly influences the final generated content. These weights are determined based on several factors, including the source's historical accuracy, the context of the information, and the model's training data. However, an AI might extract a fact from a less reliable source without recognizing its inaccuracy, leading to the dissemination of false information. This issue underscores the importance of creators being meticulous with their databases, ensuring that AI systems draw on correct and wholesome information. Even with objectively correct information on its own, a direct-to-user approach poses its own risks relating to information integrity. Users relying on AI-generated summaries lose the context of the original web pages from which the information is derived. Media literacy education emphasizes identifying bias, accuracy, and good sources, but AI systems are unable to gauge these context clues or convey them to users. This limitation results in two issues: AI is more likely to deliver false or incomplete information, and users are more likely to believe it. Beyond initial misinformation of the user, these false conceptions then infiltrate into human-backed publication; a New York lawyer presented and submitted an AI-written brief full of non-existent case references in 2023. The perceived accuracy of AI, combined with the lack of contextual background, reduces readers’ ability to critically assess the information they have easy access to.
Impact
The biggest change comes with the retooled Google search engine that is being rolled out first in the US, and slowly to different areas. The new system provides an “AI Overview,” a summary version of all of Google’s results delivered to the reader. In this way, information is delivered directly to the user without prompting them to actually visit any of the related websites. As a result, from small businesses and large journalistic corporations, these website owners lose out on significant traffic. Semrush, an analytics tool that measures website traffic from Google has already identified a downturn in site visits for websites including New York Magazine (32%), GQ (26%), and Urban Dictionary. On the other hand, user-generated content sites such as Reddit, Quora, and LinkedIn saw spikes in visitor activity. This shift may be related to users placing their faith in AI engines for “fact,” while maintaining their demand for subjective opinion. In addition, the AI summary can often pull from these sites, lending credibility and visibility to misinformation. Smaller entertainment new sites also claim that the newer Google algorithm pulls from a smaller range of sources, an especially problematic phenomenon when users are looking for news updates, where diverse sources with different and competing biases are integral to gaining a holistic view of events. Alternatively, the search engine is intelligent enough to figure out what the user wants to hear and which sources they want to hear from. Many platforms, like the New York Times, have reported on how search engines and AI-driven platforms use algorithms to tailor search results to individual users. These algorithms analyze user data, including previous searches, clicks, and interactions, to deliver results that align with the user’s interests and preferences. This worsens the echo chamber concept by reducing the chance of readers recognizing that the information they are consuming does not come from an objective source. As such a dominant player in how the majority of the public receives information, AI-tooled Google search can support incorrect or harmful ideals without beginning to challenge them at any point, which leads to a less educated and misinformed society.
The transformation of the information environment surrounding consumers has significant consequences for news organizations themselves, mainly stemming from traffic diversion from traditional news journals. The business models of these sites depend heavily on clicks, visitors, and engagement to remain viable. As platforms like Google evolve into one-stop shops for consumer inquiries, traditional news sites lose their most valuable resource: their user base. This decline in traffic directly impacts revenue, compelling publications to reduce operational costs by cutting down on staff, shortening publishing times, and prioritizing the most efficient systems. This increased reliance on AI for content generation can negatively affect the human element of journalism, diminishing job opportunities and reducing the local impact of these publications. Additionally, a consistent user base provides an essential resource through visitor data, enabling news sites to optimize content, user interfaces, and operational processes, as well as leverage that data for advertising and trading purposes. Without this data, these organizations struggle to curate content effectively, weakening their ability to serve their audience's needs and lose yet another source of incoming funds. The shift toward AI-driven content curation and the subsequent loss of direct user engagement pose a threat to the quality and diversity of journalism, undermining the role of news organizations in providing informed and balanced perspectives to the public.
Conclusion
In conclusion, while the partnership between OpenAI and News Corp represents a leap forward in integrating AI in journalism, it also represents profound challenges in the revolution of information environments. The integration of AI promises enhanced efficiency and content delivery, yet issues such as misinformation, bias, and the erosion of journalistic integrity are significant. As AI tools become more embedded in news production and distribution, striking a balance between technological advancement and safeguarding the trustworthiness of information becomes paramount. The future relies on diligent adaptation, ensuring that AI augments rather than corrupting the critical role of journalism in informing and empowering societies globally.
