We recorded the language of millions of Free4Talk conversations. English rules, as you'd expect - but the rest of the table is nothing like what most people would guess.
By Mira Sonder · TrackOnSight · 2026-07-10
Ask anyone what language Free4Talk is in and they'll say English, and they'll be right. What they won't guess is everything underneath - the shape of the table once English is set aside. We've been quietly recording it, and the order of that list is not the one most people would predict.
A word first on where these numbers come from, because it matters. TrackOnSight began watching Free4Talk in late February 2026, and everything here is drawn from what it has logged since - roughly four and a half months, across about 120 days of tracking. Free4Talk itself is far older; it has been running since 2017. So this is a snapshot of one recent stretch, not the platform's whole life, and because the scraper went dark on a handful of days, every total below is a floor rather than a ceiling. The real figures are higher. Even as a floor, the picture is striking.
Start with the giant. Across the tracking window, English rooms accounted for more than 2.6 million hours of recorded conversation, spread across some 273,000 different people. Nothing else is close. The next language on the list, Hindi, comes in around 254,000 hours - barely a tenth of English's total. On a platform built for language exchange, that lopsidedness is exactly what you'd expect: English is the reason most people are here, the common tongue everyone is reaching for, so nearly everyone passes through an English room sooner or later.
The surprise is what fills the rest of the top of the table. After English and Hindi come Arabic, at roughly 220,000 hours, and then Bengali - and Bengali alone, at around 216,000 hours, outranks French, German, and Spanish put together. Keep going down and the pattern holds: Urdu, Telugu, Tamil, Nepali, and Malayalam all post serious numbers, as do Indonesian and Vietnamese. The languages of South and Southeast Asia dominate the board. The European languages a Western learner might expect near the top - French at around 84,000 hours, German at 40,000, Spanish at 33,000, Russian at 25,000 - sit well below them.
This isn't an accident of the data; it's a portrait of who actually uses Free4Talk. The platform's centre of gravity is in India, Bangladesh, Pakistan, Indonesia, and Vietnam - mostly young learners practising English for school, for exams, for work - and the table simply reflects where those millions of people come from. If you've spent time in the rooms and felt like the whole world was there, the research backs up that impression, and so does this: the "world" of Free4Talk leans heavily South and Southeast Asian, with a strong Arabic-speaking presence alongside it.
The pairings make the same point from another angle. Free4Talk rooms carry a primary language and often a second one, and when you count the combinations, English sits at the centre of almost every popular pair. English-and-Hindi is the single most common pairing by a wide margin - over 800,000 recorded sessions - followed by English-and-Arabic, English-and-Bengali, and English-and-Urdu. Read those pairs and you can see exactly what's happening: a Hindi or Bengali or Urdu speaker in an English room, first language listed beside the one they're practising, using the familiar as a bridge to the new.
One pairing stands out for not involving English at all. Hindi-and-Urdu - two languages so close that their speakers can largely understand one another - turns up more than a hundred thousand times across both orderings. It's a quiet, human detail buried in the numbers: speakers of neighbouring tongues finding each other and talking, no English required.
There's a practical finding hiding in how rooms label their level, too. Most rooms - around 57% of all recorded sessions - are set to "Any Level," open to everyone. That's the default, and it's where the bulk of the platform lives. But among the rooms that do specify a level, the balance is surprising: "Advanced" and "Upper Advanced" rooms together make up about a quarter of all sessions, while "Beginner" and "Upper Beginner" rooms account for only around eight percent. The levelled rooms skew advanced, not beginner.
That's worth knowing if you're just starting out. If you filter for beginner rooms and find slim pickings, this is why - there simply aren't many. As a new learner, most of your practice will happen in "Any Level" rooms, shoulder to shoulder with everyone else. That isn't a problem, but it does mean the skill of reading a room and choosing the right one matters more than hunting for a label that mostly isn't there.
Is any of this shifting? Over the most recent couple of weeks the ordering barely moved - English on top, then the same South and Southeast Asian cluster, with Vietnamese, Bengali, and Indonesian ticking up slightly against the fortnight before. That's too short a window, on too young a dataset, to call a trend; treat it as a still frame, not a direction. The deeper pattern - English first, the subcontinent filling the rest - is the stable one.
What the map really tells you is what Free4Talk is for. It isn't a European café of French and Spanish dabblers; it's a vast, mostly young, mostly Asian workshop for practising English, with dozens of first languages feeding into it. If that's what you came for, you're in the right place, surrounded by more people doing exactly the same thing than any table of numbers can quite make feel real. And if you came to practise something rarer - Malayalam, Kurdish, Afrikaans, all of them real and present in the data - you're not alone either. You just have to know which door to open.