How publishers can protect their businesses from the threat of AI

The rapid rise of artificial intelligence has sent publishers scrambling to understand the potential threats it poses to their business models and how they can best react and adapt to protect their interests.

The addition of AI-generated responses in Google’s search results pages, which could result in significantly less website traffic, is causing the most consternation. Meanwhile, chatbots such as OpenAI’s ChatGPT threaten to fundamentally change how consumers search for and interact with information, potentially disrupting publishers’ relationships with their audiences.

Toolkits spoke with David Buttle, an independent consultant and former Director of Public Affairs and Platform Strategy for the Financial Times, to understand how publishers can quantify the risks AI disintermediation could pose to their business models, how to future-proof their editorial and content strategies, and how to ensure their interests are protected in licensing negotiations with AI platforms. 

Quantifying the threat of AI

For publishers hoping to mitigate the impact of AI on their businesses, the first step is understanding where and why their current content and business models could be exposed and disrupted, and where they can invest to defend themselves and insulate their businesses against AI disruption.

“That starts with an analysis of what types of content they’re producing that are dependent on Google,” Buttle said. “What kind of search terms are driving traffic to those different types of content? What’s the intent behind those search terms? And finally, how effective is AI at serving those user needs that manifest in those queries?”

With a better understanding of the traffic they stand to lose, publishers can adjust their content focuses accordingly.

“It becomes about habit formation and driving engagement when users are already on their properties. Before they think about licensing or suing, that’s where you start – by understanding and defending your current business model against this to the fullest extent that you can.”

Which publishers and content are most likely to be disrupted

Some types of content are more likely to be disrupted or replaced by AI chatbots and large language models, which means some publishers are significantly more exposed than others to AI disruption. 

“If your content isn’t distinctive and it’s on the commoditized end of the scale, there’s going to be lots of it that can be synthesized and presented by an LLM, whereas if it’s highly distinctive, then the users will have to go directly to publishers in order to find it,” Buttle said.

“Understanding which content is dependent on Google is where publishers should start. There are ways you can look at your content in relation to how reliant it is on search, but also how it might provide insulation against AI in obvious and less obvious ways. For example, you can look at your content by topic and, and what topic areas are you covering. You can also look at the format. Is it multimedia? Is it text only? Is it video images and so on? But you can also think about what function is it serving the user. The last way of thinking about your content is likely to be the most interesting when considering the utility of AI in any given use case.”

Evergreen content is directly in the firing line, Buttle added, because much of the utility it offers has already been baked into large language models. Publishers don’t have the same opportunity to take measures to protect it the way they might for news content, for example, where the value is in the instant it’s published and soon after. 

Content catering to simple low-stakes searches is also more likely to be disrupted by AI. If users don’t need a trusted source for a specific piece of information, they’re more likely to trust the response of an AI system and not need to validate it by going to a trusted media brand that they’re familiar with.

A growing emphasis on direct audience revenue

For some publishers, AI is giving rise to a model where casual readership is monetized primarily on third-party platforms, while publishers focus their efforts instead on monetizing their most engaged audiences via subscriptions, events, and other products sold directly to audiences. In that scenario, publishers with healthy licensing arrangements and robust subscription businesses may feel they’re relatively well-positioned to capitalize as AI platforms continue to gain traction. 

“The relationship between publishers’ advertising revenue and their traffic is probably a reasonably direct one. So if AI means a decline in traffic, then it probably means a decline in advertising revenue too. That could push publishers further towards direct audience revenue,” Buttle said.

“I think publishers need to do two things: One is to defend their existing business by driving engagement on owned and operated platforms. But simultaneously, they need to be thinking about what’s going to be happening over the medium to long term. I think it’s highly likely that a greater proportion of publisher revenue in the future needs to come from licensing content as an input to a user-facing service.

“Finding the right balance to maximize aggregate revenues is the challenge. Because if you go too far down the licensing route, you are going to create big cannibalization risks for your direct readers who are going to be delivering greater revenue per set of eyeballs.

“So thinking about a nascent licensing business, what red lines are you putting around that, and how are you segmenting the readers that you’re prepared to reach via a licensing business versus those that you want on owned and operated platforms and to have that direct commercial relationship with? These are the questions publishers need to be grappling with.”

Is Google crossing a line with AI overviews in search results?

Over the past 20 years, Google has slowly incorporated more of publishers’ content directly into its search results pages via rich snippets and elsewhere across its products. The addition of AI overviews could be viewed as a logical progression of that approach.

“I think from a user point of view, you can make the case that AI overviews are a natural extension, but from an intellectual property and a publisher point of view, I do think there’s a rubicon that’s been crossed,” Buttle said.

“Google is expressly creating responses for users based on content that publishers have no choice but to give it access to because otherwise they’ll be cut off from search referrals. And it’s doing that to create a product that is directly substitutional to publishers’ products. Furthermore, it’s monetizing that through advertising. So I do think philosophically there’s a Rubicon crossed in terms of its relationship with publishers, even if you can make the case that this is a natural extension of what it’s been doing in its search engine results page for some years.”

Currently, publishers shouldn’t expect to opt out of having their content used for AI training purposes and still expect Google to send traffic their way. That ultimatum could catch the attention of regulators, Buttle suggests.

“I think it’s likely that competition authorities will be looking at that because to me, it seems that Google is leveraging the dominant position it holds in the market for general search to secure an advantage in the market for large language model chatbots. I don’t understand how that would be permissible under competition regimes.”

Will Google license content for AI training purposes?

“It would need a significant mindset shift within Google because AI overviews exist within general search and it’s never crossed the line of paying for the use of content to inform general search. When I last spoke to them about it, which wasn’t that long ago, they said ‘Never say never’. But I think their philosophy around search creates barriers to that happening.”

OpenAI’s motivation for publisher licensing deals

OpenAI has been on a licensing deal streak over the past twelve months, securing partnerships with major publishers including The Financial Times, News Corp, Axel Springer, Dotdash Meredith, Vox Media, and others. Some observers have argued that those deals are largely designed to placate the industry and to help stave off regulatory scrutiny, essentially asking for forgiveness rather than permission for sucking up large chunks of publishers’ content to help train its models.

“That’s not where I am on it,” Buttle said. “I think if you look at the deals they’ve done it suggests they’re looking for genuine utility. They’ve been looking for coverage across content categories and they focused on premium providers, so I think they’re building a real thing with it. 

“I think clearly they’ve got a strategy for approaching these deals and what content they want in these deals, but I think there’s a limit to how far that will go. Over the long run, I think the value exchange that exists with search where publishers give access to their content in exchange for traffic doesn’t exist in this case, so there needs to be economics for publishers or they just won’t give access to their content. 

“A lot of it will depend on how the market for large language models develops. If it’s highly concentrated and publishers have really low bargaining power, then it doesn’t look good, but that’s not really where it is at the moment. There’s a handful of developers, which suggests that it may be better than the search market in terms of monetization.”

How publishers can protect their interests when licensing content

Publishers partnering with OpenAI so far have positioned their deals as wins for their businesses and evidence the generative AI firm fairly values access to their content, but not everyone is convinced those relationships are in publishers’ best interests in the long run.

For publishers mulling the possibility of their own AI licensing deals, Buttle suggests they should enter negotiations carefully and deliberately, and with their long-term best interests in mind.

“Reductively, publishers want to maximize their incremental revenue [with licensing deals], so that means they want to create value for the other side while limiting their downside risks. That means setting parameters around what content is accessible and what can be done with it. 

So if you’re a news publisher, you may want to put a time parameter in there that means your latest news only becomes accessible to an AI client a day or two days further down the line. What can they do with it? How long can a summary of your content be? Do you want links? Do you want citations? Do you want branding? 

And I think finally, the point to say is that we don’t know how this technology is going to develop or be adopted by consumers so there’s a lot of risk associated with these deals, especially if you’re giving all of your access to all of your content to a potentially substitutional product. If the revenue number associated with licensing isn’t near your overall revenue number now, then there’s a huge risk. So I would say you want to build in a get-out option so if the deal is cannibalizing your core revenue and isn’t making sense commercially you can exit it pretty quickly.”