Short cuts: quick reads on digital publishing

The London Book Fair has announced that it is bringing back its Market Focus programme for the first time since 2022, naming Japan as the featured country for the March 2027 fair. Taking advantage of a larger venue at Excel London, the fair is introducing a new "Day 0" lineup of pre-fair conferences on Monday, March 15, 2027, which will feature The Writers' Summit, IPG AI training, and the HP ConTEXT production conference. Additionally, seminar submissions have officially opened with a new format, requiring applicants to specify a targeted stage—such as the newly renamed "The Salon"—for their proposed sessions.

In this exit interview, outgoing Book Industry Study Group (BISG) executive director Brian O'Leary reflects on his ten-year tenure and the persistent structural hurdles facing the book publishing industry. He points out that publishing heavily relies on outdated "rules of thumb" forged by historical success, making it inherently slow to adapt to modern disruptions. Despite generating billions in revenue, the business remains highly fragmented and financially constrained compared to Big Tech giants, leaving individual publishers and trade organizations with little budget for expansive, forward-thinking research. O'Leary highlights how underfunded collective efforts—such as BISG operating on just $700,000 a year to update an entire global supply chain—prevent the industry from achieving the critical mass required to proactively solve systemic problems.

O'Leary also addresses the industry's cultural blind spots, noting a repetitive cycle of short-term decision-making regarding retail consolidation, digital book pricing, and subscription models. He emphasizes that publishing has been dangerously slow to adopt modernized infrastructure, such as the ONIX 3.0 metadata standard or author identity verification (ISNI) to combat AI-generated content. Ultimately, O'Leary warns that treating the supply chain like invisible "plumbing" until it breaks leaves publishers highly vulnerable, concluding that modern industry professionals—whether they are authors, editors, or executives—can no longer afford to ignore the broader business mechanics if publishing is to survive future disruptions.

As artificial intelligence (AI) tools become widely adopted, large language models (LLMs) are increasingly involved in decision-making processes, from hiring to content moderation. This dual adoption raises a critical question: do LLMs systematically favor content that resembles their own outputs? While prior research in computer science has identified self-preference bias, the real-world implications of this bias have not been empirically evaluated.

A new research refers to the hiring context and shows that LLMs consistently prefer resumes generated by themselves over those written by humans or produced by alternative models, even when content quality is controlled. The bias against human-written resumes is particularly substantial, with self-preference bias ranging from 67% to 82% across major commercial and open-source models.

While generative AI initially promised to democratize academic publishing by helping multilingual and under-resourced scholars overcome English-language barriers, emerging evidence reveals that it has instead introduced new structural challenges. As Brendal Aformeziem writes in an analysis, A massive increase in AI adoption among researchers from non-English-speaking countries has failed to eliminate reviewer bias, as peer reviewers simply adapt by penalizing AI-associated phrasing (such as the word "delve") as a new proxy for author identity. Furthermore, current journal policies demanding strict disclosure create an unevenly distributed "disclosure trap," disproportionately burdening early-career and Global South scientists who risk stigma or punitive enforcement without institutional support. Ultimately, because AI merely polishes prose without dismantling the underlying systemic biases of peer review, true academic equity requires structural reforms—such as double-blind reviews and tiered disclosure policies—rather than a reliance on technology alone.

In a major strategic realignment, publishing giant Springer Nature today announced the divestment of its consumer media portfolio, selling the historic US-based magazine "Scientific American" to LabX Media Group and its German counterpart "Spektrum der Wissenschaft" to publishing group GeraNova Bruckmann. The transactions—which involve brands that collectively contributed approximately €25 million to group revenues in 2025—will allow Springer Nature to focus exclusively on its core global publishing business in research, health, and education, while placing these popular science outlets into the hands of consumer media specialists.

Google Play will introduce an AI insights feature for books. Similar to Amazon’s „Ask This Book“ feature (powered by AI), readers who purchase or rent an ebook from Google Play will have the ability to request a summary or ask questions to a chatbot about the book. However, publishers and authors have the option to opt out of this feature. Sweden-based subscription service Storytel is also launching a chat interface for its books, Storytel Genie.

In its first full fiscal year since going public, McGraw Hill reported flat overall sales of $2.1 billion for the year ended March 31, 2026, but successfully turned a profit with a net income of $35.3 million by cutting costs and bouncing back from a heavy net loss the previous year. The publisher's revenue is heavily dominated by digital products, which brought in $1.43 billion—roughly 68% of total revenue—compared to $669.1 million for print format. While the K-12 sector was the only segment where print still outsold digital, higher education digital sales outpaced print ten-to-one, driving an overall 12.3% revenue jump in that market to $879 million. Looking forward, CEO Philip Moyers highlighted the company's massive digital footprint of over 100 million active users and announced plans to pilot a new "Agentic AI" precision education model to further expand their digital learning platforms.

Following a recent legal victory against the pirate website Anna’s Archive, thirteen major book and textbook publishers have filed a copyright infringement lawsuit against a new pirate site, WeLib, in the U.S. District Court for the Southern District of New York. The lawsuit alleges that WeLib copied the source code and content of Anna's Archive, illegally hosting over 43 million books and 98 million research papers to attract 80,000 active monthly users. While masquerading as a public library, publishers and AAP CEO Maria A. Pallante argue that WeLib is a commercial operation that profits from stolen material by throttling download speeds for free users and forcing paid monthly memberships disguised as "donations" that range from $7 to $90 a month for faster access.

Following similar trends at McGraw Hill, Wiley reported flat total revenue of $1.67 billion for the fiscal year ended April 30, 2026, but saw its net income skyrocket 163% to $221.6 million due to internal cost-cutting and efficient AI productivity initiatives. The publisher's growth was primarily fueled by its research division, which grew 5% to $1.13 billion behind strong journal output and $49 million in AI licensing deals with 19 corporate clients and four LLM developers. These gains helped offset a 7% decline in Wiley's learning division, which suffered from retail softness and macroeconomic headwinds. Looking ahead to fiscal 2027, President and CEO Matthew Kissner expects low- to mid-single-digit revenue growth, which will be further bolstered by the company's recent $452 million acquisition of Emerald Publishing.

In his small essay, Kevin Kelly (founding executive editor of Wired magazine) proposes the hypothesis that Large Language Models (LLMs) possess an emerging systemic bias, or "attractor," toward truth. Drawing an analogy to science, he describes truth as a vast, interdependent web of coherent facts that support one another on a global scale. Because LLMs are trained on this immense, interconnected body of knowledge, they map this coherence as a gradient where confirmed facts carry more weight. Consequently, falsehoods feel misaligned and out of place within the network, meaning that the AI naturally leans toward consensus and honesty based on the sheer scale of structural reinforcement.

While critics often point to hallucinations, Kelly argues that these errors are merely the necessary price of creativity, comparing them to human dreams and a child's imagination. Through advanced engineering and nested hierarchies of oversight, these hallucinations are increasingly managed and suppressed rather than completely eliminated. Ultimately, Kelly emphasizes that while this bias toward honesty is a remarkable technical achievement that could make AIs future beacons of reliability, it also creates a distinct dilemma: truth is not identical to moral goodness. Engineers must now find a way to temper the AI's natural urge to provide accurate information with the ethical responsibility to prevent harm.

At the recent US Book Show, Hachette CEO David Shelley strongly opposed the use of AI detection software to screen manuscript submissions, warning it creates a toxic culture of suspicion. Industry analyst Mark Williams highlights that Shelley’s caution is deeply rooted in personal experience. In 2011, Shelley famously bought and edited the crime novel The Cuckoo’s Calling based purely on its narrative merit, having no idea that the pseudonymous debut author "Robert Galbraith" was actually J.K. Rowling. Williams notes that if today's flawed AI detectors had been used back then, Rowling's highly polished, professional prose would have triggered a false positive for machine generation.

To prove the fatal flaws of AI detectors, Williams ran the opening of Rowling's novel through six popular tools; two flagged it as highly probable AI content, while actual AI-generated text sailed through as "100% human." These tools do not detect the presence of a mind; they merely measure mathematical properties like sentence uniformity ("burstiness") and word predictability ("perplexity"). Consequently, exceptional, highly edited human writing—as well as text by non-native English speakers—is routinely penalized for its precision. Williams concludes that publishers must reject these unreliable algorithms and rely instead on true editorial judgment to determine if a writer is in command of a book.

A Munich court has held Google directly liable for false claims made by its AI Overviews, striking down the tech giant’s traditional legal shield. According to the Decoder the case involved Google's AI falsely linking two publishers to fraudulent business practices by hallucinating connections that never existed in the source material. Rejecting Google's defense that search engines merely aggregate third-party links, the court ruled that because the AI synthesizes, rewrites, and structures information "in its own words," the resulting text constitutes Google’s own content.

This ruling fundamentally transforms AI operators from passive conduits into legally accountable publishers. The court dismissed Google’s argument that users bear the responsibility to fact-check sources, noting that standalone AI summaries function like news teasers under press law. By stripping away search engine immunity and weakening free-speech protections for algorithmic outputs, this landmark decision poses a massive liability threat to Google, OpenAI, and Perplexity—proving that even a minor error rate can translate into millions of actionable legal risks at scale.

Generative artificial intelligence (GenAI) can enhance open educational resources (OER), but educators need a framework for responsible use. This framework should address authorship, control, instructional intent, and pedagogical value to ensure effective integration of GenAI into OER workflows. Via Educause Review