This has been the case since the launch of ChatGPT in November 2022.
Developed by OpenAI and initially based on the GPT-3.5 large language model (LLM), later updated to the GPT-4 architecture, it spawned the release of competing products, including Gemini, Claude, Llama, Ernie, Grok and Microsoft’s launch of Copilot, initially based on OpenAI’s GPT-4.
After 24 months of near continuous hype, every publisher, however large or small, is discussing, experimenting, legislating for, and in a lot of cases building prototype products for customers, or internal productivity solutions to improve workflow or reduce time taken on labour intensive tasks. All of this frenetic activity is based on or around LLMs.
There have been some notable successes and interesting case studies (see the May / June issue of InPublishing) and expert opinion has dominated face to face and virtual conference programmes, as the media industry takes stock of the implications of AI and the best way to respond and adapt to it.
The primary challenge is moving from point and click website interfaces to conversational ones through the integration of AI, chatbots and LLM Q&A services. For many, this conundrum is simply summed up as, “Build, Buy or Blend”.
Which AI path to take...
The ‘build’ argument centres around the deployment of private LLM models that are trained using the media owner’s proprietary content and data, and is therefore tailored specifically to its documents, text and images. The ‘buy’ approach involves working with the large tech companies and buying into their LLMs for speed and scale, but at the cost of relinquishing some control of their content assets, which may be used on search engines to keep traffic with the tech giants.
This conundrum is creating a watershed moment for media companies, as whilst tech giants like Google and Microsoft race to commoditise media, progressive publishers and those with narrow and deep vertical content solutions like B2B media owners and high-end B2C content publishers are trying to take control of their own destiny by embracing their own private LLM models.
Tech platforms tell publishers they should integrate into their new search tools like Perplexity and SearchGPT for the promise of royalties on subscriptions and advertising revenue share. The stark reality is that tech platforms are dangling AI search integration deals, but the numbers tell a troubling story.
OpenAI is reportedly offering individual publishers between $1-5m in licensing funds per year. It sounds like a lot, but they’ve raised nearly $20bn. Over a five-year period, that equates to approximately 5% of their total funding going to publishers, for scraping and using an entire brand’s output, commodifying and then monetising it for their own gain. OpenAI would not exist at this scale without the publishing and media community’s work, as the NYT, and more recently, Mumsnet copyright cases will argue.
When publishers are choosing to ‘buy’, what they are buying is access to the model, for instance, GPT-4 or Perplexity APIs, at a time when open-source models like Meta’s Llama3.1 and NVIDIA’s Nemotron 70B are quickly closing the gap.
Private LLM models present a clear path for publishers to turn the tables. Data from early adopters shows publisher-owned specialised AI, built on top of private LLM models, are delivering strong results with modest investment.
These publisher services are seeing up to 15x boost in page engagement, doubling of registration rates, harvesting more first-party data more quickly and 10-15% increases in subscriptions with shorter payback periods.
Publishers already possess what they need to win: deep domain expertise, trusted reader relationships, and sophisticated content knowledge. While generalist AI tools can produce homogenised, dated content, publisher-developed specialised AI agents are proving vital because they’re built on trusted reporting readers rely on.
The best model to pursue is most likely to be the one that is more and more decentralised, specialised and privately built by publishers, for their own content and audiences. Ones that are built with their agency, under their control, and driving real engagement with readers to build tangible revenue and solid bottom line growth. The only question is, which publishers will move first to capture this advantage, and 2025 will be that year for many.
The search challenge
Related to this change is the significant challenge already being experienced in the reduction of organic search traffic as the largest engines move to conversational queries and AI-generated answers.
Search Generative Experience (SGE) was introduced by Google in May 2023 and renamed AI Overviews and launched in the US in May 2024. It was rolled out to the UK in mid-July 2024.
The implementation of generative AI in search engines has led to a reduction in organic traffic for websites, as users increasingly find answers directly within search results, diminishing the necessity to visit individual websites.
With AI answer sources generating a mere 10-15% click-through rate — a fraction of traditional search traffic, publishers are experiencing audience erosion, with increased competition for top rankings in featured snippets and ‘knowledge panels’ as well as a significant dent in paid advertising revenues and affiliate traffic, particularly for large consumer publishers.
The question of how best to rank in generative AI search interfaces is now a top priority and will be a major challenge for digital publishers in 2025.
AI Overviews value more highly content that is accurate, factual and reliable, and which is backed up by a credible source. It also prefers content which is easy to read and understand, and which demonstrates experience, expertise, authoritativeness and trustworthiness (or EEAT) which is important for both being in AI Overviews and ranking well overall. This is why content written with the aid of ChatGPT will rank less well on Google.
If audio and video is included in content or is the story itself, it is also more likely to perform better in an AI Overviews environment, as it signals to the search engine that more effort has been put into the generation of the content and thought given to the user experience.
Given that conventional SEO has been considered a key skill since its inception around 1997, this shift in thinking and best practice will provide significant challenges to digital publishers competing for fewer and fewer first page positions.
This change of emphasis, away from word and image led content, to greater demand for, and focus on, audio and video led content experiences will be a further challenge for digital publishers.
It is also believed that this change in focus will give rise to the adoption of ‘360-degree commissioning’, an established concept in television, that content should be ‘multi-platform’, designed to be integrated with data and other interactive media content, all of which is published simultaneously across desktop and mobile digital platforms.
For many traditional media owners, this is a significant skills change, and a seismic shift in content creation and distribution, but the prize is opening up new opportunities for audience growth, greater accessibility, and more diversified revenue streams.
Impact of Digital Markets Unit
The UK parliament passed the Digital Markets, Competition and Consumers (DMCC) Bill on 23 May 2024, marking a significant milestone in the evolution of competition and consumer protection laws. It aims to regulate digital markets, protect consumers, and update competition law. This change to the digital landscape will create both challenges and opportunities in 2025.
Amongst key aspects of the bill is the empowerment of the Competition and Markets Authority’s (CMA) Digital Markets Unit (DMU) with wide-ranging powers to regulate big tech, by designating the biggest digital players with ‘Strategic Market Status’ (SMS). This new regime has the power to set rules on what those firms must and must not do and to help remedy competition problems.
In essence, this will allow the DMU to force tech companies to pay for news content on their platforms, and fine them if they abuse their market positions. In addition, it should also enable publishers to be more fairly compensated for content revenue which will be a huge help to smaller independent companies. Most significantly, given the CMA’s investigation into Google ad practices, it may well sanction a trial to seek a claim of £13.6bn in damages on behalf of the UK media industry, which would be a significant change in the balance of power.
The other key aspect of the DMCC for digital publishers are the changes to subscription contracts and the tackling of so-called subscription traps, where a consumer is misled into signing up for a subscription to goods or services. This is commonly done by the promise of a free trial, a reduced rate trial or sample goods where the consumer only has to pay for postage and packaging using a credit or debit card.
These traps are estimated to cost consumers more than £1.6bn per year.
Whilst the legislation has passed through parliament, certain parts of it are not yet enacted, with the law expected to be fully in effect by spring or summer 2026, depending on secondary legislation and consultations. These include the 14-day cooling-off period for consumers.
The current law allows for pro-rata refunds, or a waiver where cancellation rights are waived if the user has started using the subscription during this 14-day period. As an example, if on day eight of the subscription, the user decided a refund was wanted, they would only get a pro-rata refund, for the remaining six days not the eight days already used. The new act’s details on this are still unclear, causing businesses to delay changes, until this is clarified.
Businesses need to prepare for process changes in the delivery of reminders and ensure their software can handle these requirements automatically. Manual processes would be impractical for large user bases. In essence, it needs to be easier to cancel or change a subscription, which includes changing the payment schedule or the length of the subscription, as well as cancellation altogether. The concern many digital publishers have is that all these reminder notices are designed to put people on notice, and it’s likely subscriptions could be cancelled, rather than letting them run on.
The number of prices and complexity of some subscription contracts many publishers have, including introductory offers, multiple year contracts or free periods, combined with the ‘consent or pay’ subscriptions that have emerged since the deprecation of third-party cookies and the introduction of cookie consent pop ups in January 2024, highlights the complexity of managing subscription changes, such as upgrading payment schedules, which could lead to cancellations and revenue loss.
This is a real concern and challenge for digital publishers in 2025 when all of these scenarios need to be catered for, tested and got ready for the planned implementation in 2026, particularly given the nature of 12- and 24-month subscription contracts. This is made worse by the current lack of clarity on the size and scale of potential fines for non-compliance. The legislation is designed to protect consumers, so corporate subscriptions taken out on behalf of business users or business users buying individual subscriptions for work claimed back on expenses are not necessarily included but further clarification will be given on this as secondary legislation is finalised.
This article was first published in InPublishing magazine. If you would like to be added to the free mailing list to receive the magazine, please register here.
