Listeners told researchers this year that they are less willing to try AI narration than they were last year, and they bought more audiobooks than ever. Both findings come from the same organization, weeks apart, and the space between them is the most useful thing an audiobook producer can read in 2026.
The Audio Publishers Association released its annual surveys on June 5, 2026. US audiobook sales revenue reached $2.43 billion in 2025, up 9% over the previous year, with publishers reporting more than 750,000 active titles, a 43% increase on 2024. Demand is not the problem. The same release reports that willingness to try an AI narrated audiobook fell from 70% in 2025 to 61% in 2026, and that AI voiced titles accounted for .03% of 2025 sales revenue.
Read as a pair, those numbers look like a verdict on synthetic narration. Read one line further and they look like something else entirely.
What the 2026 APA survey actually measured
The line that reframes the rest: only 16% of audiobook listeners have ever listened to an AI voiced audiobook.
So the 61% willingness figure is mostly an opinion about a thing people have not heard. It measures the reputation of AI narration, not the experience of it. Reputations in 2026 are shaped by voice clone scandals, chart spam and platform crackdowns, and none of that has much to do with whether a well produced non-fiction chapter is pleasant to listen to on a commute.
The method matters here too. The APA's 2026 Consumer Survey was conducted by Edison Research at SSRS in February 2026, with 1,706 spoken-word audio listeners: 1,020 who had ever listened to an audiobook and 686 who listen to podcasts or spoken-word radio but have never listened to an audiobook. It is a well built attitude survey. Attitude surveys are very good at capturing what people believe about a category and poor at predicting what they will accept once it is playing in their ears.
Stated preference and revealed preference, side by side
| What was measured | Source and date | Finding |
|---|---|---|
| US audiobook sales revenue, 2025 | APA Sales Survey, conducted by Toluna, released June 5, 2026 | $2.43 billion, up 9% year over year |
| Active audiobook titles, 2025 | Same release | More than 750,000, up 43% |
| Willingness to try AI narration | APA 2026 Consumer Survey, Edison Research at SSRS | Fell from 70% in 2025 to 61% in 2026 |
| Listeners who have ever heard an AI voiced audiobook | Same survey | 16% |
| AI voiced share of 2025 sales revenue | APA Sales Survey | .03% |
| Willingness after hearing AI multi-cast narration | Edison Research study commissioned by Spoken, reported by Jane Friedman, July 15, 2026 | Listeners willing to try it and rated it favorably |
One caution on that last row, and it is not a small one. The study was commissioned by Spoken, an AI audiobook production company, so the sponsor has a direct commercial interest in a favorable result. The reporting also sits behind a paywall, which means the headline finding is confirmable and the exact percentage shift is not. We are quoting the direction, not the magnitude, on purpose. Any post that hands you a precise before-and-after figure for that study is quoting a number it did not read.
Why the .03% revenue figure says less than it looks like
Sit with the size of that share for a second. Point-oh-three percent is a rounding error inside a $2.43 billion market.
That is not a referendum result. It is a shelf-space result. Almost every AI narrated title in that year was produced by an independent author or a small press, published into a subset of retailers, with no catalogue history, no review coverage and no bestseller placement. A share that small in a market that large tells you where the titles were allowed to appear and how recently they arrived. It does not tell you what listeners thought of them.
The same survey carries a useful reality check on listener fussiness, too. Among people who listened to an audiobook in the past year, 45% listened to a free copy on YouTube, up from 38% the year before. Plenty of those uploads are pirated and none of them are mastered to a publisher's spec. When convenience is on the table, audio purism is negotiable.
What changes when listeners hear AI narration instead of hearing about it
On July 15, 2026, Jane Friedman reported a study performed by Edison Research and commissioned by Spoken, an AI audiobook production company: audiobook listeners were willing to try AI multi-cast narration and rated it favorably.
Disclose that sponsorship every time you cite it, including to yourself. A production company paying for research into whether its product is acceptable is doing marketing as well as research, and vendor commissioned studies tend to find directionally friendly answers. What survives the discount is the shape of the result, and the shape is consistent with the APA data rather than contradicting it. Willingness measured before exposure is a proxy for reputation. Willingness measured after exposure is a proxy for the audio. Those two numbers were never going to match, and the gap between them is where producers live.
Is audio quality the bottleneck, or is distribution?
Distribution, comfortably.
Spotify for Authors states it directly: "We accept audiobooks with digital voice narration on Spotify." You select "This audiobook uses digital voice narration" at upload, and Spotify adds a short sentence to the book's description so listeners know. Named accepted providers include Google Play Books auto-narration and ElevenLabs. There is one restriction worth planning around: Spotify does not currently share audiobooks with digital voice narration to referral partners. Replace the narration with a human recording later, unselect the option, and the disclosure sentence comes off.
Then there is the channel that has not opened. Audible's self-publishing submission route, ACX, is widely reported to require human narration unless otherwise authorized, with narrator-supervised voice replicas as the exception rather than externally produced AI audio. We could not load ACX's own requirements page to quote it, because that help site renders client side, so verify it yourself before you build a release plan around it. Our ACX compliance post covers the submission specifics, and the platform by platform matrix covers who accepts what.
That is the honest resolution of the whole tension. The disclosure toggle on the platforms that accept AI narration is a sentence in a book description. The gate on the platform that does not is a hard no. Nobody is failing a listening test at the door.
What an audiobook producer should do with this in 2026
Publish where disclosure is accepted and stop waiting for consensus. Spotify, Google Play Books, and the other channels that take digital narration with a disclosure line will reach listeners this quarter. Sales there are real sales, and 58% of Americans aged 18 and over have listened to an audiobook, an estimated 157 million people by the APA's count.
Audition full chapters, not sample reels. A short sample rewards a voice with personality. A long listen rewards a voice you stop noticing. Generate an entire real chapter in the voice you are considering, then listen to it while doing something else, which is how 86% of listeners say they use audiobooks. Our narration technique guide covers what to listen for on that pass, and the audiobook production kit has the chapter checklist and mastering targets we use.
Keep the human option live. Fiction still carries the market, the largest revenue share by genre in 2025 came from General Fiction, and a title with a strong performance dependency may be worth a narrator. Deciding per project rather than per principle is the whole skill, which is what our AI voices versus human narrators breakdown is for.
And keep your rights tidy. If a channel you want later requires human narration, you want a manuscript, a pronunciation list and an approved script ready to hand to a narrator, not a rebuild. Producing locally on a flat monthly price rather than per character makes re-running a chapter cheap enough that you will actually do it. The pricing page has the current numbers, and we send the next round of survey data to the newsletter list when it lands.
The 61% will move when the 16% does. Produce the chapter that moves it.
The Vois Team