Lifestyle

One collection, 57 guides, and the half of the business that faces the other way

This section holds a single collection. Audiobook is 57 guides about listening: which app, which subscription, what a credit buys, how to cancel, what speed to run, where to find things free, and how much a narrator makes.

All of it is written from the listener side of the transaction. This page is written from the other side. Not how to get an audiobook, but how one gets to you: who licenses it to whom, on what terms, why the library copy has a queue, what a synthetic voice is allowed to be called, and which of those arrangements is law and which is simply a contract somebody wrote.

The size of the thing

The Audio Publishers Association runs an annual sales survey and a consumer survey, and published both in June 2026.

Audio Publishers Association, sales and consumer surveys announced 5 June 2026. Sales fieldwork by Toluna, consumer fieldwork by Edison Research at SSRS with 1,706 spoken-word audio listeners aged 18 and over, both completed in spring 2026.

Where people actually get them

The consumer survey asked how listeners obtained audiobooks in the past year, and the four answers are closer together than the industry conversation suggests: 49 percent bought directly from a website or app, 48 percent listened through a subscription service, 46 percent borrowed through a digital library app, and 42 percent used credits from a dedicated audiobook service.

Nearly half of American audiobook listeners borrow from a library. Which makes the terms on which a library can lend one the single most under-covered subject in this section, and sections 02 through 06 are about it.

The five largest publishers, and what each sells a library

The license terms are not secret, but they are not on any product page either. A public library system in Washington State published them alongside the retail prices, and that comparison is the most useful single document in this whole subject.

Terms as published by Timberland Regional Library, 24 April 2024. The often-quoted 26 loans figure belongs specifically to HarperCollins; the 52-loans-or-two-years version is Macmillan.

Now the prices, side by side

Roughly four times the consumer price, for a copy that stops working. A library consortium in Wisconsin puts the general rule at three to five times the consumer price under a one copy, one user license, and gives an audiobook example of $25 retail against $95 for a license expiring after 26 borrows.

What has been happening to those prices

An advocacy group that tracks library digital pricing compared 2022 with 2024 and found the two markets moving in opposite directions.

Figures from the ReadersFirst 2024 price update. Within that, one publisher moved much harder than the others: HarperCollins ebooks up 17.8 percent and its audiobooks up 17.5 percent, while two others held ebook prices flat.

Read the last two rows together. Over the same two years, consumer audiobook prices fell by more than a fifth while library audiobook prices rose. The same group states the resulting gap plainly: libraries pay almost four times more than subscription listeners for the same audiobook.

Maryland tried to require publishers to license to libraries. A federal court struck it down in four months

In 2021 Maryland passed a law requiring a publisher that offers an electronic literary product to consumers in the state to also offer it to libraries on reasonable terms. The publishers association sued, and it did not take long.

The reasoning at the injunction stage was that the state law obstructed federal copyright law. The exclusive right to distribute a work encompasses the right to refuse to license it and to decide to whom the rights holder will transfer it, and, in the court words, the state may not mandate distribution and reproduction of a copyrighted work in the face of the exclusive rights to distribution granted under section 106.

The final judgment was blunt: the Maryland Act conflicts with and is preempted by the Copyright Act, and is unconstitutional and unenforceable. The court granted the declaration and declined to add a permanent injunction on the ground that the declaration was enough.

The other route closed two years later

There was a second theory: a library that owns a print copy digitizes it and lends the scan, one at a time, keeping the total number in circulation equal to the number of physical copies it owns. That is controlled digital lending, and the Second Circuit rejected it on 4 September 2024 in Hachette Book Group v. Internet Archive.

The question and answer are in the opinion first paragraph: whether it is fair use for a nonprofit to scan copyright-protected print books in their entirety and distribute those digital copies online, in full, for free, subject to a one-to-one owned-to-loaned ratio, without authorization. The court concluded the answer is no.

The same opinion describes the market it was protecting, and the description is a useful plain statement of how the whole system works: publishers distribute ebooks in two principal ways, direct to consumer through electronic retail platforms, and to libraries through commercial distributors under a one-copy, one-user model that allows each ebook to be checked out by one patron at a time. The court also found that the library ebook lending market is thriving.

So both routes around the license are closed. A state cannot order a publisher to license, and a library cannot make its own digital copy of a book it already owns.

Five ways to publish an AI-narrated audiobook, and five different rules about telling you

Every major audiobook platform now has a synthetic narration program. None of them is governed by any law about disclosure, and each has arrived at a different answer about whether the listener is told.

Descriptions from each company own author or partner documentation. The Audible program was announced 13 May 2025; the ACX replica beta 9 July 2025; the Spotify integration 20 February 2025. Apple carries no publication date on its own pages.

The sentence Apple makes a publisher print

Apple requires the product description to open with a specific line, and it is the most informative disclosure any of the five demands: this is an Apple Books audiobook narrated by a digital voice based on a human narrator.

That last clause is doing something the others do not. It tells the buyer not merely that the voice is synthetic but that it was modeled on a real person, which is the fact that matters to anyone thinking about where a voice came from.

The consent question, which is separate from the disclosure question

The ACX replica program is the one that changed the profession, because it is narrators cloning themselves rather than a platform generating a generic voice. Its terms, as published, are that narrators always choose the projects they audition for and accept, decide per title whether to use the replica or perform live, and are paid either a per finished hour fee, a blended structure, or a royalty share. And the platform states it will not separately use a narrator voice replica for any content without their approval.

On the union side, there is no single audiobook agreement to point at. The performers union states that it has more than 90 separate agreements with audiobook publishers and producers, that those contracts provide protections against the creation or use of digital replicas without informed consent and bargaining, and that it will grieve and litigate breaches at no cost to the narrator. Its three dated artificial intelligence agreements, from January, August and October 2024, cover video games, audio advertising and a voice model company respectively. None is audiobook-specific.

And no law requires any of it

Every disclosure in the table above is a private retailer policy that could be withdrawn tomorrow. What was checked and does not apply:

• New York synthetic performer disclosure, signed December 2025 and in effect 9 June 2026, requires disclosure when an advertisement includes an AI-generated synthetic performer. An audiobook is not an advertisement.
• The federal AI Labeling Act of 2026 would cover audio and would require both a clear and conspicuous disclosure and machine-readable metadata naming the system, version and creation time. It has been introduced and referred to committee. It is a bill.
• California Assembly Bill 2602, approved 17 September 2024, makes certain digital replica clauses in personal services agreements unenforceable. It governs whether a narrator can be contracted into a voice clone. It says nothing about telling the listener.

Three things that changed in the last eighteen months and made most pages wrong

This subject moves faster than the writing about it. Three specific changes date almost everything published before them.

One. The distributor most self-publishing guides name no longer exists under that name

Findaway Voices, the distribution route that appears in essentially every guide to getting an audiobook onto multiple retailers, was wound down. Uploading to Spotify moved to a Spotify-branded author portal, and distribution to everywhere else moved to a separate company called INaudio, with the change taking effect on 1 August 2025. Spotify own support site now describes the service as previously called Findaway Voices by Spotify. Terms for books already distributed carried over unchanged.

So any page that tells an author to sign up with Findaway Voices is sending them to a brand that was retired a year ago.

Two. The royalty deal on the largest self-publishing platform was rewritten in 2026

Announced 28 April 2026 and updated 26 May 2026, with enrollment opening on the latter date. The contractual terms document carrying the new rates is version 3.1, dated 26 May 2026. The legacy model is stated to be discontinued by the end of 2026.

The mechanism changed as well as the number. Under the new model, creators earn across all plan types and consumption methods, whether the listener spends a credit or reaches the title through the included catalog. The included-catalog side runs on something called Member Value, described as the price of the listener monthly membership minus taxes and fees, divided proportionally among the titles that member engaged with based on their retail price, with the creator share then multiplied by the contractual royalty rate.

That is a subscription pool model, of the kind music has used for years, arriving in audiobooks. The narrator pay guide covers what this means for a person recording the book, which is a different question from what the rights holder receives.

Three. A purely synthetic narration may have a thinner copyright than a human one

The United States Copyright Office published the second part of its report on copyright and artificial intelligence in January 2025, and its conclusions bear directly on section 07. Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements. Prompts alone, on current generally available technology, do not provide sufficient control. But copyright does protect the original human expression in a work that also includes AI-generated material, and using AI tools to assist rather than stand in for human creativity does not affect protection.

The text of the book is unaffected either way; that is the author work. What is in question is the recording. A human narration is a performance, and a performance is a work. A machine-generated reading of the same text, with no human shaping the expressive choices, may not be. Nobody has litigated it in this context, and this page is not saying anybody is unprotected. It is saying that the layer of rights sitting on top of an AI-narrated audiobook is a genuinely open question, which is not how the platforms describe it.

Questions this section gets, with the source named