There's a particular kind of listening experience that most people have had at least once, where a voice is technically doing everything right, for example, the words are clear, the pace is fine, and the tone is neutral and professional, but nothing is landing. You find yourself listening to the same sentence twice, or realising your attention has drifted without you noticing, or you might end up finishing a chapter with only a vague sense of what it was about.
The gap between a voice that performs and a voice that communicates is something most of us can feel but few of us can easily explain. And it's the reason I'm writing this.
I'm Tom Anderson, VP of Content at Blinkist, an app that distills the key ideas from the world's best nonfiction books into focused 15-minute listens, and for roughly two years we ran an experiment with AI narration across a decent portion of our new releases. We went in with genuine conviction, and we came out having made the decision to return to exclusively human narration for all new content.
This is my attempt to explain what we learned, why the listening experience was at the heart of it, and what it means for anyone who wants to learn more but finds themselves short on time.
Blinkist distills the key ideas from the world's best nonfiction books — 9,000+ titles across 27 categories, each readable or listenable in about 15 minutes, used by 43 million people worldwide — and because audio sits at the centre of what we do, the quality of narration isn't a production detail; it's the product itself.
Most people, when they think about what makes a good narrator, think about the obvious things. They think of clear pronunciation, appropriate pacing, and a pleasant tone. And AI narration, at its best, delivers all of those things competently. That's what made this question genuinely hard to answer for a long time. The AI narration we were testing wasn't the stilted, robotic kind that sounds like a computer reading a phone menu. It was polished, consistent, and in a blind technical audit it would have scored well on every surface measure.
What it couldn't do, though, was carry a sentence. A human narrator doesn't just read the words in front of them; they understand what those words are for, and that understanding shapes everything about how they deliver them. There's a weight behind a word like finally at the end of a long build-up, an earned emphasis that tells the listener this is the point, this is what everything we have talked about was working toward. An AI system can reproduce the acoustic shape of that emphasis without the meaning behind it, so you hear the stress but you don't feel the arrival. After enough listens, that gap stops being subtle. It becomes the only thing you can hear.
Princeton neuroscientist Uri Hasson and his colleagues showed, in research published in PNAS, that when one person speaks to another their brain activity begins to mirror each other's, a phenomenon they called neural coupling, and that the stronger that coupling the better the listener understands and retains what they've heard. Communication, at a neurological level, is a synchronisation between two minds. That's what a human narrator creates, and it's what no amount of acoustic polish can replicate without one.
Blinkist didn't stumble into AI narration. We began our experiments with it deliberately, with real conviction, over roughly two years. The benefits of AI narration for Blinkist were clear; human narration takes time to schedule, record, and edit, whereas AI narration could scale with the catalogue in ways that human production simply couldn't match at speed. Let’s also be honest, AI narration is much cheaper than using voices. But there was something more than cost and logistics behind our decisions. Across the audio industry — not just at Blinkist — there was a shared belief that the technology was on an upward trajectory, that every quarter the models got more natural and every quarter the gap between synthetic and human seemed to narrow. Blinkist was part of an industry-wide experiment, and we weren't the only ones who thought it was going to work out.
Blinkist launched Audio Blinks in 2014, well before the mainstream podcast boom made audio a default mode of consumption, and we've spent more than a decade thinking about what makes a 15-minute listen land. So when the listening tests and user feedback started telling me something was off, I wanted to immediately learn more.
The internal listening tests were telling us one thing, but the broader audiobook market was telling us the same thing even more clearly, and it had been doing so for longer than we'd been paying attention. According to data from the Audio Publishers Association, only 16% of audiobook listeners have ever tried an AI-voiced title, and AI-narrated content accounts for just 0.03% of total audiobook sales revenue. More striking still, willingness to try AI narration dropped from 77% in 2023 to 70% in 2025 — moving in the wrong direction at exactly the moment the technology was supposedly getting better. Library buyers, who represent a large and considered segment of the market, show a strong and consistent preference for human narrators.
Listeners, it turned out, had already worked this out. While we were running internal tests and debating the results, they were voting with their ears, and the result of this vote was not close.
Put simply, all new Blinks are now narrated by humans — recorded, produced, and delivered by professional narrators. The AI-narrated Blinks already in the library remain available but will be labeled, so you can always see which version you're choosing (We’re doing this right now but it’s a process so it might take us a bit longer). That labeling matters to me, because trust is a product feature: if we're asking you to spend 15 minutes with a Blink, you should know exactly what you're getting.
This isn't a retreat from innovation. Blinkist has been building in the audio space since 2014, and we'll keep doing that. But there's a difference between using technology to improve a product and using it in ways that quietly degrade the experience for the people the product is built for. No AI narration, no matter how polished, can deliver the same depth that a human narrator can, and protecting that depth is what this decision is about.
The app holds 9,000+ titles across 27 categories, with around 40 new titles added every month including weekly New York Times bestsellers, and it carries a 4.8★ rating on the Apple App Store across 148,000+ ratings along with a permanent Apple Editors' Choice badge. Forty-three million people use it. Apple CEO Tim Cook, after visiting the Blinkist team in Berlin, put it this way:
"Impressed with the growing and talented team at @Blinkist. Helping everyone understand and experience some of the best non-fiction books is a great example of how a single idea from college can evolve into a solution for a problem we all share — too little time!" – Tim Cook, Apple CEO (30 Sep 2019)
The product he was describing is still the product we're building. The voices you hear in it now are human ones, and they carry every sentence they read.
The broader audiobook market has already told us something important: listeners are good at noticing when a voice is genuinely there and when it isn't, even when they can't always say exactly what the difference is. A 4.8★ rating on the Apple App Store across 148,000+ ratings, and a permanent Apple Editors' Choice badge, aren't the result of a product that gets this wrong. They're the result of one that takes it seriously. If you've ever wanted to read more but struggled to find the time, or if you've tried audio learning and found it didn't quite stick, I'd invite you to try a Blink and hear what a human narrator does with 15 minutes and a great idea. The voice on the other end will be there.
Start your free trial with Blinkist — Get access to powerful ideas from top nonfiction books and podcasts.
Tom is VP Content and Editor in Chief at Blinkist. He is from the North of England and is Blinkist’s designated Tea Master. He holds a degree in Political History and Museum Studies from Newcastle University. Tom works with Blinkist’s team of readers, helping them extract the key insights from nonfiction books. Tom’s recommended read is \nThe Black Jacobins\n by C.L.R James
A small investment, an incredible growth opportunity: Get access to powerful ideas from top nonfiction books and podcasts.

Blinkist is an app that transforms key insights from great nonfiction books into quick, easy-to-understand highlights. The reality is – everyone loves reading, but no one has the time. Blinkist fills this learning gap and makes acquiring new knowledge effortless.