Google spent this summer defending itself against authors who say it fed their books into its models without permission. On August 27 it tried a different tack: put the book in the AI, but make the reader buy it first.
The company introduced Expert Intelligence, which it calls a cross Google initiative that helps you engage with your trusted sources through Google AI products. In its debut form it lets readers take an ebook they already own on Google Play Books and add it directly into Gemini Notebook — the research tool Google renamed from NotebookLM in July — as a first-class source. Once inside, the book behaves like any other document: users can ask questions and receive responses grounded directly in that book, complete with citations back to the passage, or spin the text into Audio Overviews, Infographics, Mind Maps and quizzes.
The launch catalog is not a token gesture. Expert Intelligence went live with more than 100,000 titles from Bloomsbury, De Gruyter Brill, Johns Hopkins University Press, Macmillan Publishers, O’Reilly Media and Penguin Random House — heavily weighted toward trade nonfiction, academic and technical publishing. Google also worked with more than 15 bestselling authors, including Steven Pinker, Michael Pollan and Jennifer Wallace, to build curated Featured Notebooks that pair a book with supplementary sources and interactive artifacts.
Ownership as the enforcement layer
The mechanic that matters most is the paywall Google wired into the product. Users must own the book through Google Play Books before they can query it. Share a collaborative notebook containing an Expert Intelligence title and your collaborators get prompted to buy their own copy; until they do, the book is simply unavailable inside their notebook. Readers can check eligibility on a title’s Play Books detail page, where a Gemini Notebook entry appears under the Tools badge. To seed adoption, Google is giving away one eligible ebook to U.S. accounts via the promo code EXPERTINTELLIGENCE, while supplies last.
Google’s framing of the use cases is deliberately domestic rather than corporate: a manager brainstorming feedback strategies against Kim Scott’s Radical Candor, a reader uploading personal journal entries alongside Dr. Mary Claire Haver’s The New Menopause, a parent photographing the contents of their fridge and asking for a meal plan grounded in Michael Pollan’s Food Rules. The pitch is that people reread books and reapply them, and that a notebook is where that rereading now happens.
The publishers, notably, showed up on the record. Penguin Random House CEO Nihar Malaviya said the design is an example of how AI use cases can be developed responsibly, arguing that requiring readers to purchase a book before they can use Gemini Notebook’s AI features to engage with that work recognizes the value of books while putting meaningful guardrails around how copyrighted content is used. Authors Guild CEO Mary Rasenberger — an organization that has been among the most aggressive litigants against AI companies — called it a positive step forward in the legitimized use of books in the AI ecosystem to the benefit of readers, publishers and authors alike, adding that these kinds of arrangements should open up new audiences for books while ensuring authors and publishers retain control.
Google, for its part, is betting the feature sells copies. Expert Intelligence, wrote Google Labs editorial director Steven Johnson and product management director Will Houghteling in the announcement, will over time create content discovery and help authors reach new and more engaged readers.
Analysis
Strip away the product demo and this is a position paper on the AI copyright fight, shipped as software.
The economics are the point. Every other major grounding play in generative AI has been either scraped-and-litigated or a flat licensing check negotiated over the heads of individual creators. Expert Intelligence is neither. It routes value through a retail purchase that already has a royalty structure attached — Google Play pays 70 percent on ebooks in most markets — which means an author whose book gets pulled into a notebook is compensated by the same mechanism that pays them when someone reads it on a tablet. No new revenue-share to negotiate, no opaque per-token pool. That is a materially different proposition from a lump-sum training license, and it is why publishers were willing to be quoted.
It is also a legally convenient one. In July, Google was hit with a class action from authors and publishers alleging that titles in Google Play Books were used to train its AI models without authorization. Expert Intelligence does not resolve that case, and Google has not conceded anything in it. But it establishes a public, permissioned, per-copy alternative to training-scale ingestion — and it does so in the exact product surface, Play Books, that the lawsuit names. Retrieval-augmented grounding turns out to be the friendlier legal posture: the model is not absorbing the book, it is reading a copy the user paid for and citing it back.
The strategic move is quieter. Google is converting Play Books from a fading ebook storefront into a rights-cleared retrieval corpus — the thing every frontier lab now wants and mostly cannot buy at scale. Content-licensing deals at OpenAI and elsewhere have concentrated on news publishers. Books, with their depth and durability, have stayed largely off the table because the rights are fragmented across hundreds of thousands of individual contracts. Consumer ownership is an elegant way around that fragmentation.
What to watch
Three things. First, expansion: Google says Expert Intelligence is coming to the Gemini app and AI Mode in Search, which pushes licensed grounding from a niche research tool into surfaces used by hundreds of millions — a much harder rights and UX problem than a notebook. Second, the source types Google has already flagged: third-party subscriptions, enterprise business research reports and academic textbooks, each with its own entrenched licensing incumbents. Third, whether the 100,000-title figure grows fast. If holdout publishers stay out and the catalog stalls, the ownership model looks like a compliance gesture. If it triples by year-end, it becomes the template every other lab has to answer.
“Requiring readers to purchase a book before they can use Gemini Notebook's AI features to engage with that work recognizes the value of books while putting meaningful guardrails around how copyrighted content is used.”— Nihar Malaviya, CEO, Penguin Random House