The app Arc Search was released last month, and its approach to browsing the web is about to drastically change the business of online news. It is only the first of many AI tools that will will change how content is consumed, and it’s likely to spur commercial news outlets to shift their business models when advertising revenue disappears overnight.
As this happens, public and nonprofit media can remain an essential, critical source of factual, original reporting in a growing landscape of unreliable automatically generated content. But we will succeed only if our audiences have strong awareness of our brands across platforms and if our funding models are not heavily tied to online underwriting.
AI tools have become increasingly common over the past year. Every major internet company has released its own flavor of large language model, the technology behind these tools. All have been trained on vast sets of data and are designed to respond to any kind of question within the bounds of the model, whether or not the answer is based in reality. The algorithms that tools like ChatGPT, Bing and Gemini use cannot always cite their sources, making them unreliable for finding verifiable information. And since integrating new data takes time, many of the models suffer from a lagging response to current news. Until recently, AI was not coming for news consumption.
That changed with the debut of Arc Search. For the first time, an AI model was shipped as an app for consuming specific content rather than generating it from a massive dataset. The philosophy behind this new web browser is that the internet is not a collection of web pages to be visited but rather a source of information to be communicated.
For a general internet user, its starting point is familiar enough: a search box. However, the results of a search are something else entirely. What Arc returns is not a set of links but an AI-generated summary of those links in sections organized under meaningful headlines. To prevent the hallucinations inherent in the output of any large language model, it checks to make sure specifics in the summary, such as numbers and quotes, actually exist on one of the source pages. The end result for a user is a clean, readable, cited summary of their search results that eliminates the need to click through every link.
It’s easy to see the appeal of such an app. Commercial news outlets covered their sites in distracting ads to maximize revenue. Trying to read a news article among the popup to encourage newsletter sign-ups, the midscroll autoplaying video, and animated insertions between every paragraph has become, frankly, a nightmare. On desktop, it’s annoying but manageable. On mobile, it’s almost impossible. Consuming web content without dealing with any obnoxious ads or filler content is objectively better from many readers’ perspectives.
Whether or not Arc Search succeeds, its model is about to be copied widely. I believe that will mark the beginning of the end of ad-supported commercial journalism online, since the market for web ads will effectively vanish as more and more consumption of publisher content is an input to a model rather than to a person’s eyes. This leads to only one outcome on the commercial side, which is that the only sustainable news model will rely on subscriptions or other non-display advertising dollars.
The move to a subscription model creates two problems. First, most consumers will only subscribe to a couple of independent sources due to the cost, siloing regional reporting. Second, only larger population centers will be able to sustain a subscription base to support the work of news organizations. In the end, the market for free “news” will be increasingly served by outlets providing opinions and propaganda who don’t need to turn a profit. Reliable fact-based reporting will be available only to those who can pay for it.
For public media, the effects of this shift are less clear. As AI-powered tools spread, public media will serve an increasingly important purpose as a source for free, high-quality, unbiased content, both for human users and for client-side scrapers generating ephemeral summaries, like Arc is doing today.
This requires that we start thinking about our future funding models so that they will be diverse enough to withstand the complete loss of display ads. We also need to consider how we market our purpose and mission in multiple places, to increase touchpoints where AI is less likely to be an intermediary.
This includes things many of us are already doing, such as buys of paid physical media and interacting with our audiences on as many social platforms as we can handle. It should also include changes to our owned digital presences, especially apps. The public media app ecosystem could be so much better. In fact, it’s so bad that many stations are writing their own, or joining coalitions to do the same.
In reality, there should be one app that uses the new single sign-on technology to deliver video audio, and written content from PBS, NPR, member stations and our partners. The user experience would be better, and the marketing of benefits to the communities we serve and the value of membership would be clearer to communicate.
I also hope that we can elevate, partner with and promote other nonprofit newsrooms in our communities to ensure that quality journalism is available for everyone independent of their ability to pay. We can’t control the evolution of technology, but we can do our best to meet people where they are with whatever tools they want to use.
Jason Katterhenry is information technology director for AZPM.