PubTech Radar Scan: Issue 27
2025 - Trends, Predictions, and Prediction Markets
Welcome to the first issue of Pub Tech Radar in 2025 and 🙏 thank you for continuing to let my ramblings into your inbox. Rather than the usual scan of the latest developments in academic publishing tech here are my top 10 trends for the year ahead:
AI uncertainty: David Crotty described 2024 as a year of "uncertainty" in The Scholarly Kitchen, I think 2025 will be a year of AI uncertainty as publishers grapple with how to use, invest in, manage, and respond to AI. The result will likely be a patchwork of projects, approaches, and sentiments. I think we’ll see exciting new initiatives launch while some experimental projects, particularly early GenAI projects, are discontinued. Efforts to incorporate AI into peer review will accelerate with startups driving innovation in this space. Publishers will continue to embrace AI for marketing, advertising, and production processes, but many will struggle to balance researchers’ use of AI and research integrity issues, often resorting to cumbersome reporting requirements.
AI inequality: I expect to see widening disparities in the ability to think strategically about and capitalize on AI opportunities between larger, wealthier publishers and their smaller counterparts. I’m not sure if this is a good or a bad thing! One area where technological support seems essential is research integrity, particularly for tasks like network analysis. Beyond that, the impact of these disparities may vary. Traditional, "handcrafted" publishing done on a smaller scale is likely to create a far higher quality product compared to more efficient but automated processes, also see Trend 3 below.
Good enough: In the latest episode of Midnight at the Casablanca with Damian Pattinson, the discussion touches on how publishing has evolved over the past 20 years. For instance, the BMJ's decision to split the Editor-in-Chief and CEO roles fostered a more business-oriented culture and there’s the amusing anecdote of a subversive tie-wearing protest at PLOS in response to increasing corporatization. Journal publishing is transitioning from a quaint, cottage industry into a series of increasingly standardized industrial processes run by larger organizations. This shift toward industrialization has brought concepts like failure tolerance and the acceptance of "good enough" outputs over striving for perfection. While Elsevier might deny using AI in its production processes, a quick browse through the offerings of any production vendor reveals extensive automation, some of which, depending on your definition, is usually classified as AI. I suspect we’ll see more instances of editorial teams expressing deep dissatisfaction with "good enough" outputs in 2025, as they clash with traditional expectations of quality.
Rights rising: I think rights management concerns will be ever present in 2025. Questions around ownership, the sale and transfer of rights, the use of materials as training data, and efforts to reform copyright law will remain at the forefront. There will be tricky issues to resolve for example, can a researcher create a digital twin of themselves by uploading all their published works into an LLM without running into significant rights and permissions issues? To what extent will what a researcher can do be governed by where they are based?
Compliance challenges: I’m not going to write too much on this one but the need to comply with new legislation - such as the European Accessibility Act (EAA) directive and the EU AI Act - will be a continuing trend for 2025. On top of that, companies will face added pressure to ensure their staff, and especially their vendors, adhere to internal AI usage guidelines.
AI & Metadata: I think we’ll see more discussion about using AI to generate metadata, see Thad McIlroy’s work. I think we’ll start to see more discussion about media provenance technologies like C2PA being adopted by academic publishers. I’m much less certain about this as a trend but I hope we’ll see more discussion about how publishers could use their content, expertise, and classification systems to work with AI companies on Graph RAG (Graph RAG makes it more likely that LLM chatbot responses are contextually accurate and well-connected to real-world relationships) rather than just handing over the raw content. I hope we’ll see more initiatives like nanodash and that these initiatives are matched with practical applications by AI solution developers.
KYC-lite is coming: Following on from the STM report on Trusted Identity in Academic Publishing I think we’ll see much more talk about how Know Your Customer (KYC) approaches/solutions can be applied in academic publishing in as inclusive a way as possible.
Overwhelm, anxiety, and burn out: This trend isn’t specific to technology. There’s so much change happening within academic publishing driven by continued outsourcing, acquisitions, restructuring, and business model change. Add in process changes, increasing adoption of automation and AI, and combine with ambitious growth/revenue targets and the pressure to catch all potential fraud/problematic papers, and I think we might be facing a wave of mental health issues.
Traffic takes a hit: I think changes in discoverability tech such as AI Summaries within Google search results, general services like ChatGPT and Gemini Deep Research, and specialist AI search tools such as Undermind, Elicit, and SciSpace will start to reduce traffic to academic publishers journal websites in 2025. What do you think? Try placing a play money bet on How many academic publishers will publicly state that their referral traffic from Google or Google Scholar has decreased? [See notes at end of post about prediction markets for more info]
Sentiment for ScholarOne turns positive: Ending on a positive note, I really hope the acquisition of ScholarOne by SilverChair and the promised investment will turn a service largely seen as a slow-moving blocker into something much more positive, vibrant, and dynamic that better supports the needs of researchers and publishers.
I would love to hear what you think will happen in 2025.
Other people’s 2025 predictions that caught my attention:
Deanta: “Affordable technologies will enable smaller publishers to take back control of the publishing process while maintaining quality and reach. Technology partnerships replace traditional publishing outsourcing”
Journalism.co.uk: Predictions for journalism 2025: newsroom strategy, talent and leadership. “Lexie Kirkconnel-Kawana, CEO, IMPRESS: short-terminism will continue to plague the industry”
Hong Zhou: Navigating the Digital Frontier: How Emerging Tech Trends Are Shaping Scholarly Publishing. The advent of agentic AI may support “Workflow optimization: Monitor and optimize the publishing journey from submission to peer review to publication, predicting bottlenecks and autonomously suggesting corrective actions.”
The Bookseller:AI, collectible editions and reading for pleasure dominate 2025 book trade predictions: Paul Kelly, CEO of DK, “AI remains a transformative force, offering opportunities to enhance discoverability, streamline workflows and better connect books with audiences. However, publishers must keep authors’, illustrators’ and photographers’ interests at the forefront, balancing innovation with ethics to ensure the industry’s long-term health.”
And finally…
A personal note, if I’ve blanked you, answered a question weirdly, randomly (or more randomly than usual) I apologize! On bad days I have lost almost all of my hearing 🧏♀️. Most of the time I can get by with a single hearing aid but on bad days it’s a bit of a struggle. (Hopefully, there will be no more admin errors or lost referrals and I will make it to the top of the NHS waiting list to get things sorted out soon!) - Helen
Notes on Prediction Markets and Manifold Markets:
💡Inspired by the Hard Forkasts 2025, I thought it would be fun to experiment with using a prediction market to make forecasts. However, I quickly discovered that this is not as straightforward as it seems. I decided to try Manifold Markets, one of the few platforms that lets you create your own questions. Unfortunately, some of the platform's full features—like the ability to purchase credits for setting up multiple questions—are only accessible to U.S. residents. An even bigger challenge, though, is crafting questions that can yield clear, black-and-white answers. Trends are much easier than making concrete predictions, plus, the kinds of things you might want to bet on for a bit of fun, such as will Signals have more customers than ClearSkies by the end of 2025? would be impossible to get data on.