Babbel.

A quiet giant in a noisy app store

Someone in a cafe is staring at a menu in a language they do not speak. Their thumb hovers over a phone. Ten minutes later they order, out loud, and the waiter does not switch to English. That small, unglamorous victory is the entire point of Babbel.

It is the kind of company that does not trend. There is no owl threatening you with guilt notifications, no leaderboard turning grammar into a blood sport. Babbel sells something less viral and harder to fake: the ability to actually say the thing you needed to say. Fourteen languages. Lessons short enough to finish before the coffee goes cold. Tens of millions of subscriptions sold, mostly by people telling other people it worked.

Today Babbel operates out of Berlin, with an office in New York, under the gloriously unromantic legal name Lesson Nine GmbH. In 2025 a co-founder handed over the CEO chair, the company shipped an AI speaking coach, and it kept doing the deeply unfashionable thing it has always done: charging money, on purpose, for lessons built by human beings.

Everybody "learns a language." Almost nobody speaks one.

Here is the open secret of language education: most of it produces people who can conjugate verbs on a worksheet and freeze the instant a stranger says hello. Years of classroom French, and you can still only confidently ask where the library is. The gap between studying a language and using one is where most learners quietly give up.

In the mid-2000s that gap had a new, glaring symptom. The internet could teach you almost anything, except - reliably - how to hold a conversation in another tongue. There were dictionaries, flashcards, forums, and a great deal of optimism. What there was not was a structured, paid course that treated you like an adult who wanted to talk by the end of the week.

Four Berliners, one side conversation

The story starts, as good ones often do, with people working on something else. Markus Witte, Thomas Holl, Toine Diepstraten and Lorenz Heine were building an online music platform in Berlin. Somewhere between releases, a chat about learning Spanish turned into a more interesting question: why was there no decent way to actually learn a language online?

In 2007 they founded the company. In January 2008 it went live as a free beta, because that was the era's reflex. Then they did the thing that made everyone wince - they put up a paywall. Charging for language lessons when "free" was the entire spirit of the web was either naive or contrarian. It turned out to be the business.

The bet was simple and slightly stubborn: people will pay for something that works, and "works" means you can speak. So they designed lessons around real-life conversation, kept them to about ten or fifteen minutes, and had linguists - not algorithms, not crowds - write every one.

Short lessons, long memory

Open the app and you do not get a firehose. You get a lesson roughly the length of a coffee break, built around something you might actually say. New words come back later, on purpose, spaced out so your brain has to reach for them - the unglamorous science of spaced repetition doing its quiet work. Speech recognition listens to your pronunciation. Short podcasts and review sessions tuned to your own mistakes fill the gaps.

In September 2025 the company added Babbel Speak, an AI feature that runs you through expert-designed conversation scenarios - ordering at a cafe, booking a hotel - and nudges you with feedback. The marketing line, "from silence to speech," is unusually honest about what learners are afraid of. Notably, Babbel let AI play the patient conversation partner while keeping humans in charge of writing the actual curriculum.

The numbers behind the claim

Skeptics are right to ask whether any of this actually works, because most language apps quietly measure success in downloads rather than conversations. Babbel's pitch is more falsifiable: the company reports that 92% of users improved their proficiency in two months, and it backs completion with CEFR-aligned certificates rather than confetti.

That restraint is the part competitors rarely copy. While the category chased growth at any cost, Babbel raised roughly $34 million across its life and leaned on paying customers instead of investor patience. Partners like Scottish Equity Partners and Reed Elsevier Ventures wrote the checks; the customers kept the lights on.

Mutual understanding, one lesson at a time

Strip away the app-store screenshots and Babbel's stated goal is almost old-fashioned: mutual understanding through language. Not points, not badges - the ability to connect with another human in their own words. That framing explains the company's whole posture, including its refusal to hand the curriculum over to a machine.

When the rest of the industry sprinted toward AI as a labor-saving replacement, Babbel took the slower position that technology should enhance human-designed teaching, not stand in for it. It is a less exciting headline. It is also the reason the lessons still read like they were written by someone who has actually been misunderstood in a foreign country.

The case against forgetting how to talk

Real-time translation earbuds are getting good. You can already point a phone at a sign and watch the words rearrange themselves into English. So why learn a language at all? Because translation gets you the meaning and loses the moment - the joke, the rapport, the slight thrill on a stranger's face when you meet them halfway. Babbel is betting that the more frictionless machine translation becomes, the more people will want the human version on purpose.

For business, the stakes are plainer. Teams are spread across borders, and "we'll just use the translation tool" only goes so far before something important gets lost. Babbel for Business sells the unglamorous fix: colleagues who can actually talk to each other.

Back to the cafe. The menu is still in a language you did not grow up with. But now the order comes out of your own mouth, the waiter nods, and for a moment the gap between studying a language and speaking one simply is not there. That gap was the whole problem. Closing it, ten minutes at a time, is the whole company.