What makes a listener give up on an audiobook? On November 5 RNIB published a YouGov poll that asked exactly that, and the answers read like a casting brief. 42% of UK adults would stop listening over a celebrity narrator they are not a fan of. 30% would stop if the narrator's voice was very different from how they imagined the character. 26% would stop over a strong accent. Only 16% said nothing about the narrator would make them stop.
None of those reasons is about how a voice was made. Each is about whether the voice is the person on the page, and that is the part of an audiobook an author decides.
What the poll found
RNIB commissioned the poll for the 90th anniversary of its Talking Books library, a free service for anyone with a sight condition. It holds 56,000 titles, which were borrowed more than a million times last year. The first Talking Books went out on 7 November 1935, on shellac discs, to soldiers blinded in the First World War, as Voice Over Herald's write-up recounts. To mark the anniversary, RNIB President Victoria Smurfit and 19 fellow actors from Rivals recorded excerpts from their favourite books.
| What would make UK adults stop listening | Share |
|---|---|
| A celebrity narrator they are not a fan of | 42% |
| A voice very different from how they imagined the character | 30% |
| A narrator with a strong accent | 26% |
| Nothing about the narrator would make them stop | 16% |
Source: YouGov for RNIB, press release of November 5, 2026. RNIB did not publish the sample size or fieldwork dates. Of five accent options covering England, Scotland, Northern Ireland, Ireland and Wales, the English accent was the most popular.
Ready to try it on your own book?
Start with 5,000words free →Every reason to quit is a casting failure
Read the three reasons as a director would.
A voice unlike the character (30%). A reader of fiction arrives with the cast already in their head, and the author put it there: the age in a line of description, the region in a turn of phrase, the temper in how a character speaks to others. When the voice contradicts the page, three in ten listeners leave. The fix is not a better voice in the abstract. It is a voice that matches what the book already says about the person.
A celebrity they are not a fan of (42%). The largest number is about a voice chosen for who it is rather than who it plays. A famous name is a bet on the listener's taste in that person. A voice cast from the text is a bet on the book.
A strong accent (26%). Accent is a casting decision too, and it belongs to characters more than to narrators. A Glaswegian dock worker should sound like one. A narrator carrying twelve hours of description in a strong accent is a different choice, and a quarter of UK listeners say they would stop over it.
Put the three together and the format question answers itself. In a single-narrator edition, one voice plays every character, so a single mismatch between the narrator and a part runs through the whole book. In a cast, each part is its own casting decision, which can be heard, judged and changed on its own. That does not make a cast right for every book, and some books are better with one voice. It does make the 30% a solvable problem.
Listeners judge what they hear
The poll also asked about AI narration as a category, and most respondents said they would not choose it. Asked in the abstract, people answer about a label. What they do when they hear a production is a different measurement. In a blinded test we covered in September, 1,005 US fiction listeners heard the same thriller as a professional single narrator or as an AI multi-cast edition, and rated it before anyone told them which was which. The multi-cast edition scored higher on favourability, 61% to 53%. Willingness to listen to an AI audiobook rose from 31% before hearing an excerpt to 65% after. The study was commissioned by an AI audiobook company, so read its margins as a signal, but the direction matches RNIB's own numbers: listeners reward voices that fit the characters.
What an author does with this
Cast from the page. Before any voice is chosen, write down how the book describes each main character: age, where they are from, how they talk. That is the voice your readers already hear, and it is the brief.
Separate the narrator from the characters. Decide the narrator's register on its own, for long stretches of description, and give characters their accents where the book gives them one.
Audition before you commit. Hear each main character's voice against the scene where they first speak. If you hear the person differently, change the voice before the rest of the book is built on it.
On Midsummerr, every speaking part is cast from how the book describes the character, and Martin, the AI audio director in the studio, recasts on a note. Tell it "the captain should sound older and more weathered" and it designs a new voice and plays you the sample before anything changes; then it re-records only that character's lines, in the chapters you choose. Midsummerr is the only audiobook platform with an AI audio director, and How to recast a character's voice shows the whole flow. To hear what one voice per part sounds like, Five Children and It streams free, with a narrator and 51 character voices, from the five children down to the Psammead.
The honest version
This is one poll, and RNIB has not published its sample size or fieldwork dates, so treat the figures as a single survey rather than a census of UK listeners. It measures stated preferences, what people say would make them stop, and it is a UK poll; the accent result in particular is a home-market finding.
It is also an anniversary poll from a library that exists for access. For a blind or partially sighted reader, an audiobook is not a format preference; it is how the book reaches them. The release states no position on AI narration.
What the numbers do settle is useful for anyone making an audiobook. Most listeners have a reason they would quit a narrator, and the reasons are about fit: the wrong person, the wrong voice for the character, the wrong accent for the part. Those are casting decisions, and an author can make them well.
