Artificial intelligence, and the opportunities and hazards it presents for creators and members of the publishing industry, was a recurring theme at a number of panels on the first two days of the Bologna Children's Book Fair. On Monday, Finnish illustrator Pirita Tolvanen opened her masterclass, “Drawing Reality: Authenticity and New Conventions in the AI Era,” by arguing that nonfiction picture book illustrators should be far more transparent about their research, and that the AI era makes doing so essential as a means of fighting disinformation. Drawing on a 2022 survey of Finnish illustrators, she showed that the most rigorous practitioners visit locations, interview experts, and sketch on-site, while others do little more than a final fact-check. The gap, she said, shows in the work.

Tolvanen walked through examples of illustrators who are doing it right, using back matter that includes dated sketchbook pages and on-site photographs, author’s notes that explain exactly where a picture departs from scientific reality, and what she calls “inventory illustrations”—detailed kit and equipment spreads—whose accuracy depends entirely on the artist having handled the real objects. For historical and biographical subjects, she raised the question of when a recognizable likeness actually matters, and when motion and context can carry the weight instead.

Her larger argument was pointed: in a world where AI can generate convincing evidence for events that never happened, children’s book makers have a new responsibility to show their work. “It’s not just about evaluating if something is true,” she said. “You have to evaluate the evidence you have.” Back matter, done well, is one answer.

The discussion about AI continued on Tuesday with Bologna Book Plus’s AI Summit, which opened with Nadim Sadek, founder and CEO of Shimmr AI and author of two bestselling guides on AI—Shimmer, Don’t Shake and Quiver, Don’t Quake—arguing that AI has the potential to supercharge creativity for virtually everyone. Sadek described AI as a “superfriend”: a tool that expands rather than diminishes human creative capacity.

That optimism was tested in the summit’s midday panel, “Predictive Publishing: How Data Is Rewriting Decision-Making,” where three AI startup founders argued that the industry’s acquisitions process is structurally broken, leaving diverse new authors buried in unread submission pipelines while editors lack the tools to process what they receive. Rishiraj Chowdhury of Quantifiction cited one publisher with 150,000 titles sitting unread—“the hard labor of love of 150,000 authors who never saw the light of day”—as evidence of the scale of the problem. Gavin Marcus of Storywise said that his system is built to highlight manuscripts rather than filter them, asking not “is this book good or bad?” but “is this book good for you?” and matching submissions to the specific taste profiles of individual editors and imprints. Arsim Shillova of Libraro said that a competition his platform ran over the course of two months drew 7,000 entries, attracted 10,000 readers, and yielded 30 commercially viable titles backed by 3.2 million reader behavior data points.

The founders pushed back on the concern that algorithmic tools inevitably produce homogenized, commercial-driven content, with Chowdhury arguing that AI trained across millions of manuscripts removes individual editorial bias and respects the distinct creative “signature” of each genre. Skepticism came from the audience nonetheless, with one questioner noting that WPP’s synthetic audience models, built with far greater resources, have yet to demonstrate consistent predictive accuracy. Chowdhury acknowledged the limits. Even the best predictions will be 80% accuracy,” he said, adding that the best human editors still overlook books like Harry Potter. Marcus closed with a note of caution, arguing that distribution remains the industry’s most underappreciated variable and that the real opportunity for AI lies not in forecasting hits but in helping the people making acquisition decisions make better ones.

Mary McAveney, president and CEO of Abrams Books, closed out the summit and offered perhaps its most measured perspective. She described Abrams’s approach to AI as deliberately cautious. She emphasized the importance of building internal sandboxes for low-stakes experimentation, establishing an AI steering committee, and moving at a pace the organization can absorb. “There are no shortcuts,” she said, whether for commercial fiction or the publisher’s high-end illustrated titles. She drew a firm line between operational uses of AI, including data analysis, business intelligence, coding tasks, and the creative process, the latter of which she said remains off-limits at Abrams. She cited the global reach of Jeff Kinney’s Diary of a Wimpy Kid franchise, published in 72 languages, as evidence that what makes books connect with readers across cultures is craft, not efficiency. “It’s about getting the right illustration, the right translation, the right slang,” she said.

One unexpected finding from Abrams’s internal AI rollout, McAveney said, has been that younger employees, not veterans, have shown the most resistance. They are driven not only by job anxiety, but by a broader unease about what the technology means for humanity. Her closing message to attendees of the summit was caution against binary thinking. “The magic that happens when a child connects with a picture book for the first time,” she said, “is something I don’t want to cede to AI.”