Case Study For Babbel
Case study — a working prototype built for Babbel's AI Content Engine

Babbel needs tons of new lessons. AI can write them fast. But can we trust what it writes?

THE PROBLEM - WHAT IS CURRENTLY HAPPENING

People run out of lessons on Babbel. You could finish a whole level in a few weeks, and then, nothing. No more lessons.

Babbel also doesn't teach popular languages like Japanese or Korean yet. Why? Because writing lessons by hand takes forever — about 35 hours of work for just one small lesson pack.

AI can write lessons in seconds. So why not let AI write everything, all the time?

Because AI still makes mistakes. It can get grammar wrong. It can use a joke that makes no sense in another language. Sometimes it just gets facts wrong. If those mistakes go straight out to millions of learners, people stop trusting the app.

This actually happened to Babbel's biggest competitor. They let AI write tons of content, fast — and learners started complaining the lessons felt sloppy and wrong.

My solution

Think about how a spam filter works. Your email doesn't go straight to your inbox — it gets checked first. Safe stuff goes through. Sketchy stuff gets flagged so you can look at it yourself.

Greenlight works the same way, but for lessons. Every lesson the AI writes gets checked automatically and sorted into 3 piles:

Green
It's good, goes live right away
Yellow
Mostly fine, but a real teacher double-checks it
Red
Something risky — a real teacher must fully check it

AI does the fast part. Humans stay in charge of anything risky.

Greenlight content review flowchart

This shows up in two ways

Topic Packs

Extra lesson packs on things you actually care about, like "German for job interviews" or "Spanish for football fans."

Palpack

Matches you with someone else learning the exact same language at the exact same level, so you always have a study buddy and a new question to talk about each week.

The numbers behind this

35 hours
How long it takes to hand-write just one small lesson pack today
85%
How many AI-written lessons passed a human check in Babbel's own early tests
14 vs 40+
Languages Babbel teaches vs. its biggest competitor — because hand-writing lessons is slow
76%
How far a competitor's stock has fallen from its 2025 peak, as trust in AI-written content became a real business risk

These numbers are public information about Babbel and its competitors, verified as of July 2026. this "case study" itself is a prototype, not a live product.

What this could be worth

This is math, not a result — This case study is a prototype, so nothing here has actually happened yet. But here's what the numbers say could happen if it works.

~$900,000
Extra revenue if people stay subscribed just 1 month longer, across 100,000 learners
~$100,000/year
Extra revenue if just 1 more person out of 100 renews, across 100,000 subscribers

Estimates based on Babbel's own published numbers — not internal data, and not a promise.

View working prototype