For every news publisher out there who has been worried about the existential risk posed by AI-powered search, the past couple of weeks have brought positive news.

It is no longer a foregone conclusion that Google’s search generative experience (SGE), which summarises results instead of sending readers to news sites, will knock out the top of publishers’ sales funnels. The reason: Google is considering putting in place a paywall for AI-powered results. Users will need to pay to receive this summarisation service and will still have the option of clicking on links if they do not want to pay.

Ah, look at that: Big Tech joins the ranks of the many readers of this newsletter who have already put their faith in paywalls. And for much the same reason: It costs money to provide information that adds value for users. 

In fact, it costs Google 10 times more to generate a list of summarised results than it does to return a typical, old-fashioned page of results with links, according to my colleague INMA Researcher-in-Residence Grzegorz Piechota.

Google is not alone in coming up with ways to cover its expenses. Perplexity, another generative AI platform, is planning to help fund the costs of its AI-powered search results by placing ads against them. (Of course, if you read my March 14 newsletter, you already saw this coming.) 

A few publishers have already hedged their bets on search by signing licensing agreements with GenAI platforms to provide training material for their language models. They will receive not only millions of dollars in revenue but also preferential treatment in AI-powered search results, with links back to their sites.

But for every other news publisher out there: Surely greater friction for users on search engines for intelligent results has to be positive. How many readers at the top of your funnel are willing to pay for information or love the user experience with ads? Would they rather pay Google up front for saving them some time — or would they rather click through to your site in the hope of finding what they are looking for without spending a penny? 

Before you heave a sigh of relief, consider this: This decision by search engines to monetise GenAI search is hardly unalloyed good news.

It is a reminder that GenAI is an immature technology. Even the providers of it are still evaluating the costs of compute and of running queries at scale. You will recall ChatGPT had entirely free access when it was publicly launched, and the folks at OpenAI were shocked when the bot gained a million eager users within five days, with each query costing OpenAI a few cents to run. Two months after its launch, OpenAI started charging for a premium tier and limited the number of users accessing its free product during peak hours.

Across the world, every news publisher that I have spoken to over the past three months is using GenAI tools. This means the industry now faces the risk of escalating costs — because whichever GenAI platform is in use, it is almost certainly going to be more expensive. 

Are we ready for this? It is likely that costs will eventually decline as the technology matures, and competition among GenAI providers could also keep a lid on prices — but we may need to ride this out.

If you’d like to subscribe to my bi-weekly newsletter, INMA members can do so here.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .