An AI reading assistant is now built into ebooks and audiobooks, answering reader questions about a character, a term or a chapter without leaving the page. For authors, the terms on which those answers are given differ sharply between platforms, and some tools offer no way to refuse.
TechBullion describes a reading assistant as a chatbot linked to one specific book. When a reader highlights a passage or taps an ask button, it answers from that book's contents rather than from a general training set, and many hide spoilers and quote the relevant section so the reader can check the answer against the page. The publication treats that distinction as the point: a general chatbot works from training data and can invent things, while a reading assistant stays grounded in a single title.
How the major tools handle permission varies. Amazon's experimental Ask this Book reached the US iOS app on 11 December 2025, offering spoiler-free answers about plot and characters when a passage is selected. According to TechBullion, the feature is always on and authors and publishers cannot opt out; Amazon calls it an extension of search that needs no licence.
Google Play Books announced Book Insights in July 2026, described as a spoiler-free reading companion that authors can turn off. In August, Google's Expert Intelligence delivered more than 100,000 eligible ebooks to Gemini Notebook, where owners get cited answers, quizzes and audio overviews.
Other products take different routes. Audible's Ask a Question beta, TechBullion reports, answers questions about out-of-copyright classics while narration continues. ElevenReader's voice chatbot lets a reader ask a book questions aloud and hear replies in the narrator's voice; ElevenLabs reports a 24% rise in listening time and 78% completion for users with five or more sessions. Sinai. AI works only from public domain or approved titles and pays rights holders, while iChatbook gives explanations without quoting the text directly.
The Authors Guild addressed the Amazon question in a statement dated 23 December 2025. TechBullion reports that the Guild called Ask this Book unlicensed, said it earned authors nothing and offered no opt-out, and, given Amazon's share of ebook sales, described it as a dangerous precedent. The Guild asked for a permissioned, paid model.
What the chat window gives writers, and takes
TechBullion frames the in-book chat window as both a tool and a worry for authors. Reader questions, when a platform shares them, show which chapters confuse people, which ideas land and where readers stop, and can shape a next edition, a talk or a course. The same window can put words in an author's mouth. The publication's test of whether it helps is who runs the assistant and on what terms.
For publishers, the article describes two pressures arriving together. Interactive editions could add value across a catalogue, with audio, recaps, study questions and guided prompts, and Google already turns eligible titles into quizzes and audio overviews. Used without permission, the same technology could take that value away. TechBullion argues nonfiction, education and professional lists stand to gain most because their readers are looking to apply what they read.
The data picture also changes. A retail sales report tells a publisher how many copies moved, not how far anyone read. A branded reading experience with opt-in consent gives a publisher what TechBullion calls actionable first-party data, including finished works and reader questions, and the publication treats that reader data as valuable intellectual property.
On contracts, TechBullion notes that most publishing agreements were drawn up before in-book chatbots existed. The terms it says publishers need include opt-in by title, payment by title or per use, limits on quoting, and a ban on training outside models with the text.
Praxis AI Media, which the article names as an example of the permission-based route, takes titles from authors or publishers, does not sell reader data, and leaves existing distribution including Amazon in place while adding a direct reader relationship. Its publisher tiers support unlimited titles.
What separates the products, in TechBullion's account, is not the question answering but the consent behind it. Some tools ask authors first and pay them. Some stay with public domain titles. Some, including ElevenReader and Gemini Notebook, accept uploaded files. The article's closing claim is that the assistant most likely to succeed is the one that asked first.
Sources
Every quotation and figure in this article is checked against the source reporting linked above before publication. How we report