AI Search and the Threat to Online News

Generative AI-powered search engines are on the rise, and online news publishers are starting to worry this trend will impact traffic to the very content informing AI search results. (If this seems reminiscent of the threat social media news linking posed a decade ago, it should.)

Analysis by the International News Media Organization (INMA) found that 29% of traffic to news sites would be in jeopardy if AI search continues to grow.

In this article, we examine the threat of AI search and look at what news publishers can do to defend their traffic, content, and revenue.

Traffic at risk

The major problem with AI search is that instead of a traditional search model, where results appear as a series of source links organized by relevance, AI search generates results as sentences. For example, if someone searched for the history of post-modern literature, an AI search engine would return a paragraph summarizing the information instead of offering a list of sources to explore. This is problematic for media outlets who rely on direct and search traffic to generate revenue, especially now that Google — which, according to INMA, represented 93% of the global search market in May 2023 — has entered the ring.

Google Search AI and the generative AI powering it, “Search Generative Experience” (SGE), was announced during Google’s I/O conference in May. Unlike AI chatbots like ChatGPT and Google Bard — which are trained on existing data sets (ChatGPT, for example, can’t write about anything past 2021, making it obsolete for generating news) — SGE uses AI modelling to aggregate information drawn from the internet and summarizes that information in response to a search query. As CNBC reports, this development escalates the longstanding concern that the “snippets” Google includes in search results prevent users from clicking through to the source: “Publishers have long worried Google repurposes their verbatim content in snippets on its own website, but now Google is using advanced machine learning models that scrape large parts of the web to ‘train’ the software to spit out human-like text and responses.”

“‘You could essentially call this the Wikipedia-ization of a lot of information,’” Bryan Goldberg, CEO of Bustle Digital Group, told The New York Times . “‘You’re bringing together Wikipedia-style answers to an infinite number of questions, and that’s just going to nuke many corners of the open web.’”

Like Wikipedia, SGE cites its sources, but, as Washington Post analyst Geoffry Fowler notes, these source links do not necessarily encourage traffic. “When Google’s SGE answers a question, it includes corroboration: prominent links to several of its sources along the left side,” explains Fowler. “There are two ways to view this: It could save me a click and having to slog through a site filled with extraneous information. But it could also mean I never go to that other site to discover something new or an important bit of context.”

Nieman Lab takes it a step further, suggesting SGE “could significantly decrease the traffic that Google sends to publishers’ sites, as more people get what they need right from the Google search page instead. They could also do some damage to the affiliate revenue that publishers derive from product recommendations.”

Beyond AI search’s threat to traffic — and revenue — publishers are also grappling with a series of other issues, including AI’s well-documented propensity for spreading misinformation, the rise of AI journalism, and questions of consent and compensation surrounding the content used to train large language models (LLMs).

Publishers fight back

With so much on the line, publishers are beginning to fight back against AI search engines.

Vanity Fair reports, “The trade group Digital Content Next—whose members include the Times, NBCUniversal News Group, The Washington Post, BBC News, Axel Springer, Bloomberg, Condé Nast (the parent company of Vanity Fair), and numerous other major media organizations—recently issued a set of ‘principles for development and governance of generative AI.’ Among these: ‘Publishers are entitled to negotiate for and receive fair compensation for use of their IP,’ and ‘copyright laws protect content creators from the unlicensed use of their content.’”

Other publishers have dedicated internal resources to combat AI. According to The New York Times, “Vice Media has created a task force in recent months to examine its own approach,” and The Washington Post “appointed a deputy business editor to lead an internal group looking at A.I.’s impact.” News Corp — which owns major publications like The New York Post and The Wall Street Journal — reportedly chose to appeal straight to the source and is “in talks with ‘a couple of companies’ about the use of its content.”

Developers have responded with less than convincing assurances that AI will not destroy online news. Microsoft’s head of Bing said, “directing users to click through to publishers was ‘a top goal,’” while a representative for Google stated, “the company was ‘deeply committed to supporting a healthy and vibrant news ecosystem’ and would put a priority on sending traffic.”

Reducing 'search dependency'

Only time will tell if publishers and developers can find solutions to the issues of content use, appropriate compensation, misinformation, and declining traffic — but in the meantime, the online news ecosystem should prepare to adapt to a world of AI-powered search.

What’s New in Publishing (WNIP) recommends publishers reduce their “reliance on search to drive traffic and build first-party data.” Slate agrees, suggesting “Email, podcasts, and other subscription media” as viable alternatives to traffic-based revenue.

Cultivating a community of engaged, loyal audience members will be critical to this approach, especially as Google continues experimenting with different applications of AI — the latest of which is an AI journalism tool “that can generate article text and headlines,” according to CNN. Though Google claims “it is looking to partner with news outlets on the AI tool’s use in newsrooms,” only time will tell if that promise of collaboration comes true. In the meantime, it’s become more important than ever for media outlets to build strong relationships with their audience and prioritize the kind of high-quality content AI can’t simulate.