If the book industry doesn't decide how AI gets used in books, somebody else will.
Tech companies, platforms, and vendors are writing the early rules about how books get described, surfaced, and sold in the AI era. They are not the people who built this industry, and they are not the people who answer to authors when contract terms change. They are here to capitalize on what the rest of us spent careers building.
This is the part where the book industry has to show up for itself. And one of the organizations doing that work is the Book Industry Study Group.
Key Takeaways
- 46% of individuals and 48% of organizations in publishing reported using AI tools for work, according to the 2025 BISG and BookNet Canada AI survey.
- 98% of respondents reported at least one concern or pain point with AI only 2% had none.
- Fewer than 30% of organizations have an official AI policy, and only 27% use closed or enterprise AI environments that protect proprietary data and IP.
- The biggest finding is a policy gap: AI tools are already inside publishing houses, but the frameworks meant to govern authorship, copyright, and editorial standards have not yet been written for most of the industry.
Fifty Years of the Industry Studying Itself
BISG was conceived at the Book Manufacturers Institute's fall meeting in November 1975, when a cross-section of publishers and manufacturers recognized the industry needed better research, operating statistics, and shared standards. BISG was formally incorporated as a not-for-profit in February 1976. Its first report, on book industry information needs, was published two months later.
In 2026, BISG turns 50.
Across those five decades, the organization has done what its name promises. It studies the book industry. It convenes the supply chain. It builds the standards (BISAC subject codes and ONIX metadata among them) that the rest of the industry runs on. BISG published the Book Industry Trends report for more than two decades starting in the 1980s. It partnered with the Association of American Publishers on BookStats. It has run accessibility surveys, consumer surveys, and supply-chain studies that shaped how this industry understands itself.
BISG is what it looks like when the book industry decides to study and govern itself, instead of waiting for somebody else to do it.
Why Did BISG Launch an AI Working Group?
In December 2023, BISG launched a dedicated AI Working Group, reporting to its Workflow Committee. The mandate is exactly what you'd want it to be:
- Collect and evaluate information on AI applications used in book publishing.
- Lead and facilitate discussions on both effective and concerning uses of AI.
- Propose best practices that balance innovation with safeguards for IP, privacy, and bias.
- Publish findings and recommendations for the industry.
The group includes publishers, librarians, retailers, service providers, and standards experts. The people in the room have decades of publishing experience between them. That matters. AI questions in publishing are not, at their heart, technology questions. They are questions about contracts, authorship, supply-chain trust, metadata integrity, and reader experience. The people best equipped to answer them are people who have spent careers inside the industry, looking at it from every angle.
The most consequential AI decisions in book publishing won't be made in a server room; they'll be made by people who understand what a contract, a copyright, and a reader actually are.
What Did the First Industry-Wide AI Survey Find?
In summer 2025, BISG and BookNet Canada ran the first dedicated, industry-wide survey on AI use across the North American book industry. The response was significant enough that BISG noted it as one of the top three response rates of any BISG survey in the past decade.
An industry frequently described as cautious or slow-moving showed up in unusual numbers to be counted on this question. People in this business want to be heard on AI.
The survey gathered 559 total responses, with 521 from the United States and Canada. Brooke Horn, BISG's Membership and Operations Manager, presented the data at the Human-Aligned & Ethical AI in Publishing Summit. What follows is the picture that data painted.
Adoption Is Outpacing Comfort
Three numbers tell the central story:
- 46% of individuals reported using AI tools for work.
- 48% of organizations reported using AI tools.
- 98% of respondents reported at least one concern or pain point with AI. Only 2% had none.
Read those three together and you have the tension this industry has to resolve. Nearly half of publishing is using these tools. Almost everyone in publishing has serious reservations about how they're being used. As Brooke put it directly in her session: The industry is adopting AI a lot faster than it is becoming comfortable with it.
Publishing has reached the point where the question is no longer whether to use AI, but whether the industry will govern that use before someone else writes the rules.
Where Is AI Showing Up in Book Publishing?
For individual contributors, the most common AI use cases are practical:
Administrative and operational tasks (meeting notes, email, calendar) led the list.
- 21% reported using AI for data analysis
- 19% for marketing
- 15% for metadata and title optimization
At the organizational level, administrative work still led, with marketing, data analysis, and editorial close behind. Least common across both groups: rights and licensing management, QA testing, customer service, reader engagement, and translation.
AI is being absorbed first into the work publishing teams already wished they had more help with. The repetitive, time-consuming, administrative layer that has quietly grown around every role in the industry. That maps directly to what we believe AI should be doing for publishing: craft stays human, drudgery goes to AI. (You can read more about that philosophy in our other Next Chapter AI writing on human-aligned workflows.)
The Policy Gap Is the Real Story
Less than 30% of organizations in the survey reported having an official AI policy. Only 27% are using closed or enterprise AI environments, which are the kinds of systems that actually protect proprietary data and intellectual property.
Among publishers specifically, responses split roughly evenly across three buckets: have a policy, don't have a policy, working on one. Libraries showed even less structure. Service providers had the highest rate of formal policies.
This is where the data should land hardest with publishing leadership. The tools are already in the building. The work is already being done with them. The policy framework that should govern how that happens has, for most of the industry, not been written yet. That's the gap where author trust, IP, and editorial standards are most exposed, and it's the gap that vendors and platforms will happily fill in their own contracts and terms of service if publishing doesn't fill it first.
An AI policy is the difference between a publisher setting its own terms and a publisher inheriting someone else's.
What Concerns Does Publishing Hold About AI?
The most frequently reported concerns are the ones publishing has been talking about in hallways and back-channels for two years:
- Inadequate controls around the use of copyrighted material (86%).
- AI hallucinations (84%).
- AI-generated books flooding platforms (81%).
- Inaccurate, false, or biased training data (79%).
These aren't fringe positions. They are dominant concerns held by the overwhelming majority of an industry that is, in the same breath, using AI in daily work. That contradiction is exactly what BISG's data makes legible. It is what publishing has to resolve with its own values and its own framework, before the answers get written for it.
We're Proud to Have BISG as a Summit Partner
BISG is a Supporting Partner of the Human-Aligned & Ethical AI in Publishing Summit because their work is foundational to what we're trying to do.
Next Chapter AI helps publishing teams build literacy, policy, and human-aligned workflows. BISG sets the standards, gathers the data, and convenes the working groups that make industry-wide answers possible. Our work serves individual organizations. Their work serves the system. Both are necessary. (If you're starting from scratch, our Next Chapter AI policy and literacy resources are a good place to begin.)
The 2026 BISG AI survey is coming. The next standards conversations are already underway. If you care about how AI gets used in publishing (and you should, because the decisions being made now will shape the next decade), joining BISG is one of the most direct things you can do. Membership is open to publishers, librarians, retailers, service providers, and individual professionals across the supply chain.
This is the table the book industry built for itself. It's the one we want deciding what comes next.
FAQ: BISG and AI in Publishing
What is BISG?
The Book Industry Study Group is the leading book trade association in North America for standards, best practices, and supply-chain collaboration. Founded in 1976, BISG turns 50 in 2026. Its members include publishers, retailers, libraries, service providers, and individual professionals.
Is the 2025 AI survey the first one BISG has done on AI?
Yes. It is the first dedicated, industry-wide AI survey from BISG and BookNet Canada. The AI Working Group launched in December 2023 and conducted the survey in summer 2025. A 2026 follow-up survey is coming.
How many people in publishing are actually using AI?
Per the BISG survey, 46% of individuals and 48% of organizations are using AI tools. Service providers lead at 82% individual adoption, well ahead of publishers (43%) and libraries (42%).
How many publishing organizations have a formal AI policy?
Less than 30% of organizations in the survey reported having an official AI policy. Only 27% are using closed or enterprise AI environments. This is one of the largest gaps the data exposed.
Where can I read the full BISG AI survey report?
BISG and BookNet Canada published the full white paper through bisg.org. Members get access to the deepest cuts of the data, including segment-level breakdowns and qualitative responses.






