Alternatives to Audible for people who listen to learn

For years I listened to audiobooks and podcasts on my commute. I would finish an hour with the feeling of having understood something, and find little of it there when I tried to explain it later. A monologue cannot be stopped to ask why.

I build Understand, one of the tools at the end of this post, so weigh my conclusion with that in mind.

Most people who search for an alternative to Audible want the same books for less money, so I start there. Prices are for the US and were checked on 3 October 2026.

Cheaper and free audiobook services

Audible charges $14.95 a month for Premium Plus, which includes one credit for a book you keep. In March 2026 it added a Standard plan at $8.99: one book a month, yours for as long as you keep paying.

If money is the reason you are leaving, start with Libby. If you already pay for Spotify Premium, you have 15 hours a month you may not be using, about one long book.

A 12-hour book costs the same 12 hours on every one of these services. If you listen to learn, those hours are the larger cost.

Summary apps

Blinkist sells 15-minute summaries of nonfiction, over 9,000 across books and podcasts, for $15.99 a month or $99.99 a year. Headway does the same with about 2,500 titles and bills in four-week blocks ($19.99 for the first block, $39.98 after). Shortform charges $24 a month, or $16.42 a month billed yearly, for longer book guides with audio narration.

A summary keeps a book's conclusions and drops its argument. The objections the author answers and the steps you might have rejected go first, because they take the most room. What remains is a list of claims you can agree with and cannot defend.

I use summaries to decide whether a book deserves ten hours. I have never understood a hard idea from one.

Read-aloud tools

Some apps read articles aloud, along with PDFs and ebooks you already have. I used Speechify for this. Its free tier has ten robotic voices and tops out at 1.5x; Premium is $29 a month, 60% less on the yearly plan, with natural voices and speeds up to 5x.

ElevenReader, from ElevenLabs, has the better free offer: 10 hours of text-to-audio a month at no cost, and unlimited listening on Ultra at $11 a month ($8.25 billed yearly). Readwise Reader reads your saved articles, PDFs and EPUBs aloud on iOS, Android and the web, as part of a $12.99 subscription ($9.99 billed yearly); it needs a connection because the speech is generated on request. On an iPhone, Safari will read a supported page for free: open the page menu and tap Listen to Page.

Pocket used to belong in this list. Mozilla shut it down on 8 July 2025.

These tools gave me more to listen to and left the original problem where it was. An article read by a synthetic voice is still a monologue. Speechify's Premium page now lists a "Voice AI Assistant" you can talk to "about anything, including any website or book"; I have not tested how it behaves in the middle of a reading, so I will not judge it here.

AI audio made from your own sources

Google renamed NotebookLM to Gemini Notebook on 16 July 2026. Its best-known feature is the Audio Overview: you upload sources, and two AI hosts discuss them in the style of a podcast.

Audio Overviews have an interactive mode. You join the conversation, ask a question, the hosts answer from your sources and then pick up where they left off. That is the closest thing in this post to what I wanted. Google's help page lists its limits: English only, a beta in the mobile app, and listeners of a shared link cannot interact. Usage is free up to a compute quota that refreshes every five hours, with paid plans raising it.

An Audio Overview is a summary with two voices. The hosts tell you what your paper says and never read it to you. For a source whose exact wording matters (a proof, a contract), you are hearing a paraphrase by a model and trusting it.

The startup version of this idea did not survive. Huxe, built by Raiza Martin, Jason Spielman and Stephen Hughes after they left the NotebookLM team, generated personal podcasts from a prompt. It announced it was winding down on 22 May 2026, a day after Spotify released a similar personal podcast feature.

ChatGPT voice

ChatGPT's voice mode is the tool I used most to ask questions out loud. According to OpenAI's help page, the Live voice can listen and speak at the same time, can use web search and memory, and gives Plus subscribers 3 hours per rolling 24 hours (free accounts get limited access to a smaller model). You can interrupt it mid-sentence, and it is good at answering a why.

It did not read to me. When I asked it to read an article, I got a summary of the article. I asked how much of the article it could see, and it said a high-level summary or the first few sentences. The separate Read Aloud button plays back messages ChatGPT has already written, so reading a document that way means getting ChatGPT to print the document first; a Tom's Guide writer who uses it for books reports that it stops to ask "Do you want me to keep reading?".

So I had Speechify for the text and ChatGPT for the questions, in two apps that knew nothing of each other. Asking about a paragraph meant leaving one app and describing the paragraph to the other.

The evidence on listening and learning

The evidence on audio as a medium is mixed. Rogowsky, Calhoun and Tallal (2016) randomly assigned 91 adults to hear, read, or both hear and read the preface and a chapter of Laura Hillenbrand's Unbroken. They found no statistically significant difference in comprehension, either immediately or two weeks later. One chapter of narrative nonfiction is a narrow test, and Daniel and Woody (2010) got a different result in a psychology classroom: students who listened to a podcast of a primary source scored lower on a quiz than students who read it. The students had preferred the podcast until they took the quiz.

The stronger evidence concerns what the learner does. Freeman and colleagues (2014) pooled 225 studies of undergraduate science, engineering and math courses. Exam scores were 0.47 standard deviations higher under active learning, and students in traditional lectures were 1.5 times more likely to fail (33.8% against 21.8%).

Roediger and Karpicke (2006) had students either reread a short passage or try to recall it. Five minutes later the rereaders remembered more, 81% against 75%. A week later the order had flipped: 56% for those who had recalled, 42% for those who had reread. In their second experiment the students who only reread were the most confident they would remember the passage, and a week later they remembered the least.

Deslauriers and colleagues (2019) measured both learning and the feeling of learning in Harvard's introductory physics course, with the same students taught some topics by polished lecture and others by active exercises. From the abstract: "Students in active classrooms learned more (as would be expected based on prior research), but their perception of learning, while positive, was lower than that of their peers in passive environments." I recognize my commute in that sentence.

Michelene Chi's ICAP hypothesis (Chi and Wylie, 2014) orders these results. It predicts that learning rises from passive (receiving) to active (manipulating, such as rewinding or highlighting) to constructive (producing something the material did not contain, such as an explanation in your own words) to interactive (constructing in dialogue with someone who responds). Listening to an audiobook at 2x is the first mode.

None of these studies tested commuters with audiobooks. Freeman and Deslauriers studied college science classrooms, Roediger and Karpicke used passages of under 300 words, and ICAP is a hypothesis supported by lab and classroom studies of note taking and self-explanation. Nobody has run a trial of Understand either. I am extrapolating.

Conjecture and criticism

Karl Popper called the common picture of learning the bucket theory of the mind: knowledge exists out there and pours into us through the senses. He argued in Objective Knowledge (1972) that this never happens. David Deutsch makes the same case in The Beginning of Infinity: all knowledge is created by conjecture and criticism, including the knowledge in a single head that we call understanding. I have written about this before in the context of LLMs.

When a narrator explains natural selection, the explanation does not transfer. I hear words and guess at what they mean. My guess may be wrong, and the only way to find out is to criticize it: ask a question, or restate the idea and see whether it survives. A monologue removes every one of those steps. The guesses pile up unchecked, and a fluent narrator makes unchecked guesses feel like comprehension.

Deslauriers's authors attribute their result to the fluency of a good lecture and to novices misjudging their own learning. I think Popper explains why fluency misleads: it removes the moments of confusion in which a learner would have noticed a wrong guess.

Understand

Understand is my attempt to put the text and the questions in one place. As a Speechify alternative it adds the questions, and as a NotebookLM alternative it reads the source itself. It reads an article or PDF aloud word for word, one sentence at a time, with the current sentence highlighted on screen. When you start speaking, it stops. You ask why, it answers, and the reading resumes from the sentence where you interrupted. Spoken commands (faster, slower, skip, go back, pause) work without a tap, and speed runs from 0.75x to 2x.

It also asks questions. When you open a saved article to talk about it, the tutor gives the gist in two sentences and then asks you a question about the first idea that is new to you, before explaining anything. It keeps an outline of the concepts you have shown you understand, which you can read and correct, and it marks an idea as something to build on only after you have succeeded with it in two different contexts. Hearing a document read counts for nothing in that outline.

Material gets in through a saved list. A bookmark button in the app takes a pasted link, a Chrome extension saves the page you are on, the iOS share sheet saves from Safari or any app that shares a link, and X bookmarks sync in once you connect your account. You can upload a PDF of up to 25 MB directly.

It has no catalog of narrated audiobooks and no human narrators; the voice is synthetic, and a book has to arrive as a PDF with real text in it, because scans are refused. There is no Android app. It needs a connection the whole time, so it is useless on a flight. It is free for now, on the web and on iOS through TestFlight.

Reasons to stay with Audible

A novel read by a good narrator is a performance, and nobody wants to interrupt a performance to ask why. For fiction and anything else you listen to for pleasure, Audible or Libby is the right tool and nothing in this post replaces it. The same goes for long drives without signal and for anyone on Android.

For the book you are listening to because you want to be able to use its ideas, try a test before your next credit renews. Stop the audio after a chapter and explain its main argument out loud. If you can, keep listening. I could not, which is why I built something I can interrupt.