Media and data companies are fighting to protect their turf as generative artificial intelligence (AI) tools begin to take hold among consumers and businesses looking for information.
Those in the news and proprietary data industries are now faced with the challenge of protecting their existing revenue streams and working out how to either fight back against AI or embrace it. After the initial burst of interest in chatbots such as ChatGPT, which scrape information and serve it up to users in a simple format (often stripped of context and accuracy), some media giants are already negotiating to sell their data to the big tech companies that increasingly dominate the AI industry. Others are looking to partner with them and build their own chatbots. “Ultimately people are looking for answers. If you can provide them more reliably than everyone else, then you will be alright," said Amy Webb, a futurist and founder of Future Today Institute.
This isn't the first time an emergent technology has created a problem for media executives. Most newspapers were left behind by the emergence of the internet, and have had to forge new revenue streams after advertising shifted to social media platforms and aggregators such as Google. Meanwhile, the advertisers that embraced digital fastest have been the most successful over the past decade.
“Not a single news organisation was prepared for their business model to be upended [by the internet], I just remember a lot of stubbornness from executives,” said Webb. “I don’t think journalism jobs are dead, but the current atmosphere looks a lot like it did in the late 1990s and early 2000s.”
Newspapers used to be the primary place people went to find information. Now, for people aged 18–24 in the UK, social media apps are the most common place to find news, according to YouGov. Today, 40 per cent of this age group get their news from social media apps, 39 per cent from news websites and just 8 per cent buy newspapers.
Accordingly, advertising revenue shifted from the newspapers to the tech platforms. In 2000, global newspaper advertising revenue peaked at $150bn, but it has now fallen to just $35bn. To put that into perspective, last year Meta (US:META) made $113bn.
The newspapers that have thrived are those that produced high-quality specialist content and placed it behind a paywall, such as the New York Times (US:NYT) and the Financial Times, the IC's sister publication. The New York Times's increase in subscribers has pushed up its revenue 38 per cent since 2017 to $2.3bn. FactSet consensus forecasts that its free cash flow will hit $266mn next year.
"Some companies responded to the internet by trying to produce as much content as possible and drive traffic to their websites," said Webb. "But we are seeing now with the streaming platforms like Netflix that ultimately consumers want to pay for good quality."
This strategy is not dissimilar to the business intelligence companies Relx (REL) and GlobalData (DATA). Unlike newspapers, they aren't consumer businesses, but they have both had success selling specialist data and analytics packaged in an easily consumable way. This insight has made them very profitable, with both companies generating historical operating profit margins in excess of 20 per cent.
But even that approach could be undermined by AI if they don't react quickly enough.
All the world's a page
The concern is that AI tools will remove the need to have someone filter and package information – be it a newspaper, an analytics company or a marketing agency. “AI is going to completely [cut out] everyone in the middle,” said Webb. “You can ask it anything, from its opinion on current affairs to what the best type of shampoo is and it happens instantly."
So far, the popularity of generative AI is spreading even faster than the internet did in the 1990s. ChatGPT only launched seven months ago, yet 26 per cent of 16- to 75-year-olds in the UK have used a generative AI tool, according to a report from consultancy Deloitte.
Seemingly, the major newspapers are trying to learn from their mistakes from the past by reacting more quickly to this new technology. One way to tackle this problem is for the media companies to sign agreements with the tech platforms, effectively selling their insights and journalism to be incorporated into the chatbots, rather than trying to compete directly.
AI companies OpenAI (which developed ChatGPT), Google, Microsoft and Adobe (US:ADBE) have met with executives of media companies – such as the New York Times, The Guardian and News Corp – to discuss the prospect of paying them for their content, as reported in the Financial Times.
“Media’s collective intellectual property (IP) is under threat and for which we should argue vociferously for compensation,” said News Corp chief executive Robert Thomson at the INMA conference. “[AI] is designed so the reader will never visit a journalism website, thus fatally undermining that journalism.” OpenAI has already agreed to hand grants to local US news services, as well as giving them access to its technology. This followed a deal with Associated Press to give OpenAI access to news content going back to 1985.
Meanwhile, the B2B analytics companies are going down a slightly different route. Relx is building its own chatbot by licensing foundational models from big tech companies and then incorporating its proprietary data. Relx's legal business, LexisNexis, has a lot of case data that can be integrated into a chatbot. However, unlike generalised chatbots such as ChatGPT, Relx’s product will be able to cite its sources.
Chief financial officer Nick Luff said Relx was working with multiple different large language model providers and had “hundreds of software engineers and data scientists” working to build new products. “Since the turn of the year, generative AI has become a priority for us and many customers have asked us specifically to build this product,” said Luff.
It is the quality of Relx's data that positions it well to create a useful generative AI tool for its customers.
Similarly, GlobalData's trove of data puts it in a position to build something its customers are willing to pay a premium for. Customers sign subscription deals with GlobalData to receive high-quality data on industries from life sciences to media, technology and economics. Its data is clearly highly valued as last year it was able to push through price rises, which increased its adjusted operating margin by 2 percentage points to 36 per cent.
Webb recommends the newspapers follow a similar build-your-own model. “They should be building their own systems so a reader could come to, for example, the FT and ask it a question and then receive a response from the collective knowledge of all the FT journalists ever,” she suggested. “Newspapers gave up on building good search functions, meaning most subscribers would use Google to find articles on their own websites, it was madness.”
Currently, chatbots are, supposedly, trained on public information such as Wikipedia, Twitter and Reddit. “Twitter is a circus and Reddit is mostly males between the ages of 18 and 40; it is hardly the most diverse set of training data they have,” said Inma Martinez, chair of the Global Partnership on Artificial Intelligence steering committee. “I still believe that the good writers with the salt of the earth in them will be hard to copy, they just need to find a way to protect their work.”
To build internal chatbots, media companies can use open-source algorithms such as Hugging Face or Meta’s large language model, known as LLaMa. Meta’s model has been open to academics but soon it will be made available for public companies to build their own models on top of it, according to the Financial Times.
The advantage Relx and GlobalData have over newspapers is a simple one: they are high-margin businesses with lots of cash to use for innovation. For five of the past six years, Relx’s operating margin has exceeded 25 per cent while its free cash flow has regularly been above £2bn.
GlobalData, while much smaller than Relx, has also seen consistent growth in its free cash flow, from under £20mn in 2018 to £61mn last year. There are also financial success stories in the press, but investment returns similar to the above companies are not so common. Reach (RCH), the owner of the Daily Mirror, Daily Express and Daily Star, has a return on equity (ROE) of below 10 per cent while Relx and GlobalData sit around 45 per cent on this measure, using 2022 numbers. Reach's profits are already plunging, as it reported this week.
A simple chatbot will not flip this ROE but Webb's maxim about getting customers the right information at the right time rings true. With management teams highly sensitive to missing the next great technological shift, keeping that in mind might steer them clear of both committing too heavily to AI or missing the boat altogether.
