If 2024 wasn’t quite the year of the AI revolution for media, it was certainly the year of talking about the AI revolution.
But amid all the grand talk about how utterly AI will change everything, from the hype-purveyors in tech to the often more gloomy voices in media, the key battles for AI in 2025 will revolve more around the nitty gritty of political maneuvering, savvy (or not) dealmaking and actually effective (or not) technical implementation.
Taking a look at the actual deals signed and reported in 2024 (table below these article) gives you some idea of the state of play and the direction of travel so far. (Acknowledgements to Digiday and Press Gazette).
A look down the table shows us a few things.
First, almost all are in one way or another about publishers licensing their content to AI companies, predominantly to OpenAI, one of its biggest funders, Microsoft, or smaller AI firms focused more narrowly on services that involve delivering the content that publishers create. They, for the most part, involve that data being used to train models, and for surfacing that content in response to some sort of query, with attribution.
While we don’t know the monetary value of most of the deals, what has been reported suggests that the payments are significant, but far from transformative.
For instance, the biggest number reported, the $250mn deal between OpenAI and News Corp (the original leader – remember – in those now-expiring payments to publishers by Meta and Google), provides over five years a figure that amounts to just 10% of the media business’s quarterly revenue. Revenue-sharing deals with the likes of Perplexity and ProRata are potentially intriguing, but are only likely to be a big deal if the services they offer capture consumers en masse – and that will come with a whole host of other problems.
On one level, these relationships still look like a good deal. After all, for years the arrangement swith the likes of Google and Meta/ Facebook were that publisher content would be used by those companies to drive their business models (and huge profits), while the primary kickback came in the form of traffic that publishers had to monetise themselves.
On the other, though, services like ChatGPT or Google’s Gemini are now putting even more space between audiences and the media organisations which create the content used to capture them. And, indeed, the number of court cases from various media organisations against AI companies from the likes of seemingly AI-sceptical New York Times (or even those who have struck deals elsewhere, such as News Corp) shows media companies are keenly aware that AI has the clear potential to damage their business.
Nevertheless, as initial terms of engagement, it makes sense, says James Ball, a journalist who has written extensively about the internet and its impact on our information system: “It’s probably a good transitional move: once the court cases shake out, copyright is unlikely to prevent AI companies training on articles, so being able to monetise now and to build commercial relationships will be an advantage.”
Bell adds: “The clincher will be in the lobbying game to set the ground rules that will form the basis of the long-term financial relationship between AI companies and the media. It’s tricky because something will have to give: in theory there is nothing in law stopping AI ‘learning’ facts from the media without compensation, but it breaks the old deal between search/social and news outlets. Without a new deal, it’s going to be very hard for a lot of the media to survive.”
Negotiating those future deals will, therefore, likely be in part a political question. Whatever you think of the specific solutions they’ve implemented, countries such as Canada, Australia, France and Spain have shown a willingness to step in and try to transfer some of the value tech companies have extracted from news and other media back into the organisations that create it. However, governments are also extremely keen to get in on an expected AI boom by appealing to the big beasts in AI and the investment that their exciting expansion seems to promise.
Balancing these new relationships with AI companies against ongoing lobbying efforts to ensure that the money doesn’t stop flowing – potentially through regulation – will be tricky. There’s a chance that third-party marketplaces might play a part, such as Created by Humans, a startup from the founder of Scribd that is currently trying to ease copyright deals between book publishers and AI companies. Our money, however, is on a combination of government intervention, public opinion and negotiation leading to some sort of arrangement beyond the current deals. Whether that will end up working for the news industry, though, is very much up for grabs.
One thing that’s clear is that the news media probably shouldn’t bet on the ongoing largesse of tech companies. They should know by now that these companies always have their own agendas, quite distinct from anything to do with the dissemination of news.
Recent history, such as the infamous pivot to video, driven largely by Facebook, is instructive. Publishers fell over themselves to create the video content that Facebook owner Meta wanted, in return for cash payments, only to see the company eventually begin to deprioritise news altogether. If there’s one takeaway from that whole debacle, it should be that the temporary alignment of media industry goals with the tech platforms may see the money roll in for a while – but it’s no guarantor of long-term stability or payments. The fact that many of the companies which most aggressively went all-in on video barely exist any more should underline the importance of learning the lesson.
Of course, the above list of deals also highlights the fact that media companies are – at least on the face of it – expecting to get something more than just money back from being friends with the AI industry. Most of the content licensing deals also talk about some form of collaboration, on things like advertising targeting, chat services or even vaguer new projects.
The fact is that actually working out how to use AI to improve their businesses is going to be vital for media businesses in 2025 and beyond, but it may prove almost as tricky as the larger negotiations with the AI industry. Already some of the forays into AI services have backfired, with error-strewn, artificially-generated content causing a backlash all the way back in 2023.
Proposals such as the LA Times owner Patrick Soon-Shiong’s suggestion of automatic bias meters and AI-generated alternate versions for news articles also seem at best ill-advised. But that doesn’t mean there aren’t ways that partnerships with AI firms (and, indeed, in-house efforts) can be effective if chosen and executed carefully.
Journalist Bell is clear-sighted: “In theory financial results, sports results etc, are all well-suited to it but the risks in finance of errors are high and defamation liability etc complex, so I wouldn’t be surprised if publishers are a bit more cautious here. I think it’s notable that the early adverts we saw a year or so ago for AI-powered reporters asked to produce 20+ stories a day have disappeared and not been taken up more.”
Let’s be positive.
Where news publishers really will see the benefits from AI is far less about the content they produce and much more about their ability to monetise their brands by improving how they interact with audiences and customers. Ad tech is going to embrace AI – whether they like it or not – and, while there are plenty of potential pitfalls from misfiring systems serving targeted and tailored ads, being involved in those developments (as the likes of Vox and Dotdash Meredith appear to be in their respective AI deals) seem like a necessity.
The other key area is likely to be ‘marketing and customer relationship management’ for subscription businesses. There are obvious uses to which a greater degree of automation might actually improve those services, not least in tackling (some of) the news business’s notoriously bad reputation for basic customer service. And the ability to more effectively personalise content discovery and marketing messages may well become a boon to subscription sales.
Almost every customer-facing business is looking at using AI to boost effectiveness in these areas. But having something that AI companies actually want – content – might just make it easier for the news media to build AI into its back-end systems effectively and affordably.
In many ways, the era of AI looks set to be more a step-change than a revolution for media in the 21st Century, simply because the central question remains the same. How do you navigate a complex, inter-dependent relationship with a tech industry that has inserted itself in between you and your audience? You have to hope the media business has learned at least some of the lessons of the past two decades, because the advent of AI is only making it even more existential.
| Media | Tech | Deal | Annual revenue |
| Le Monde, Prisa | OpenAI | Creation of news summaries in ChatGPT with attribution, use of OpenAI tech in building products | |
| Financial Times | OpenAI | Use of content in training data, use of summaries in ChatGPT with attribution | $5-10mn |
| Axel Springer | Microsoft | Develop products in “advertising, content and cloud computing” and create “chat-based experiences”. | |
| Dotdash Meredith | OpenAI | Content licensing and use of OpenAI tech in Dotdash ad platform. | > $16mn |
| Informa | Microsoft | Access to data unitl 2027 | > $10mn in Y1 |
| News Corp | OpenAI | Content licensing for 5 years | > $250m |
| The Atlantic | OpenAI | Content licensing, accessto OpenAI tech to provide feedback and “new experiemental site”. | |
| Vox | OpenAI | Content licensing and development of new products and services, including enhancing Vox 1st party data platform Forte | |
| Time | OpenAI | Content licensing | |
| Time, Der Spiegel, Fortune, Entrepreneur, The Texas Tribune | Perplexity | Revenue share from content surfaced in Perplexity | |
| Condé Nast | OpenAI | Multi-year content licensing | |
| FT, Reuters, Axel Springer, Hearst, USA Today Network | Microsoft | Content licensing in Co-pilot | |
| Hearst | OpenAI | Content licensing | |
| Reuters | Meta | Content licensing for MetaAI Chatbot | |
| DMG media, Sky News, Guardian | ProRata | Content licensing for use in response to queries in ProRata | |
| Blavity, Gear Patrol, The Independent, Lee Enterprises, Los Angeles Times, MediaLab, DPReview, Mexico News Daily, Minkabu Infonoid, NewsPicks, Prisa Media, RTL Germany, Adweek and World History Encyclopedia. | Revenue share from content surfaced in Perplexity | ||
| Future | OpenAI | Contente licensing, use of OpenAI products in Future sales, marketing and editorial functions | |
| Lee Enterprises | ProRata | Content licensing for use in response to queries in ProRata, development of “real time advertising solutions” and introduction of OpenAI tech to Lee outlets. |
