Some years ago I started paying attention to something small at government bank counters. The people behind the desk knew their work — the forms, the rules, the exceptions, the things that only come from doing a job for years. What slowed them down was the keyboard. A field that should have taken four seconds took twenty. A correction meant starting the line again.
I saw the same thing at other government offices. Competent people, held up by the one part of the job nobody had ever taught them. It was not a knowledge problem and it was certainly not an attitude problem. It was a motor skill that everyone had assumed would arrive on its own.
That is what eventually made me build TypeLords, and it is why I now have data on this rather than only an impression. The data is worse than the impression was.
- Education policy names coding and digital literacy. Typing is assumed, never taught.
- Across 4,830 graded typing tests on TypeLords, 82.4% fall below 35 WPM — the speed SSC CHSL requires for clerical posts.
- 75.8% of typists miss that bar even at their single best attempt.
- These are people who deliberately sought out a typing platform. The motivated ones.
- AI has changed what we type, not whether we need to.
What I kept noticing at the counter
The reason the bank counter stayed with me is that typing is usually discussed as a school skill — something children should pick up alongside computer classes, and then presumably stop needing to think about. What I was watching was the opposite. These were adults, years into stable careers, for whom the keyboard was a daily tax on otherwise competent work.
It shows up in ways that are easy to misread. A queue that moves slowly looks like understaffing. A colleague who avoids taking notes in a meeting looks disengaged. Someone who gives a short reply to a detailed email looks brusque. Often the actual cause is that producing text is expensive for them, so they produce less of it.
That is the part I find genuinely worth arguing about. Typing speed is not really about speed. It is about whether the keyboard is quietly shaping how much a person is willing to write, correct, record or explain.
The assumption sitting inside the policy
Read the National Education Policy 2020 and you will find digital literacy treated seriously and repeatedly. Coding is named directly as a skill schools should introduce. Technology runs through the document — for access, for teacher training, for bridging language gaps, for reaching students who cannot reach a classroom.
Typing is not mentioned. Neither, in fairness, would most people expect it to be. It feels too basic to legislate, in the way that holding a pencil correctly feels too basic to legislate. We have built our thinking around the outputs of digital fluency and assumed the input layer takes care of itself.
I do not think this was a mistake so much as a blind spot, and it is a very old one. Typewriting used to be taught deliberately, as a trade, with an instructor and a syllabus and a certificate at the end. Somewhere in the shift to personal computers we decided the skill had become self-evident. Children would absorb it. Adults would manage.
Mostly, they have managed. That is precisely what makes the gap hard to see — a hunt-and-peck typist is not failing at anything visible. They are just slower than they would otherwise be, permanently, at the thing they do most.
What the data actually shows
TypeLords has recorded more than 12,500 typing sessions, of which 4,830 are graded timed tests — the closest thing we have to exam conditions. Around 94% of that activity comes from India. I ran the numbers expecting them to be middling. They are not middling.
The distribution peaks between 20 and 30 words per minute. Only 17.6% of attempts reach 35. Fewer than one in ten reach 40. Below two per cent reach 50, which is roughly what a commercial data entry role would ask for.
The obvious objection is that a single test is a bad day, and that people should be judged at their best. So I checked that too, taking each typist's single fastest attempt across everything they have ever done here.
Three in four still miss it. The median personal best across 980 typists is 27 words per minute — eight short of a clerical requirement, on the single best run of their lives.
Why thirty-five words a minute matters
The 35 WPM line is not arbitrary. It is the typing requirement for Lower Division Clerk and Junior Secretariat Assistant posts under SSC CHSL — around 10,500 key depressions per hour, typed over ten minutes. The Hindi requirement is 30 WPM, roughly 9,000 key depressions. Data Entry Operator posts run a separate skill test measured in key depressions rather than words.
Put plainly: the gap between where a typical motivated learner actually is and where a government clerical job needs them to be is about eight to ten words a minute. That is not an unbridgeable gap. It is a few months of deliberate practice. But it is also not a gap that closes on its own, and nothing in a student's schooling is currently designed to close it.
This is where my bank counter observation and the exam data meet. They are the same desks. Someone clears the typing test at 35 words a minute, joins the service, and then spends the next thirty years working at a speed that was only ever the minimum. Competitive typing exams score differently from casual tests, which is worth understanding separately if you are preparing for one.
The practice illusion
There is a second finding in the data that I did not expect, and it has a direct bearing on how people prepare.
Of the 392 typists who have done both free practice and a graded test, 65% were slower on the test. The median drop is modest — about 6.5% — but the direction is nearly universal. Something about the result counting makes people slower.
Pair that with a second habit. When people quote their typing speed, they quote their best run. Across typists here with at least five attempts, a personal best sits on average around 18% above that same person's typical score. Not somebody else's score — their own.
Those two effects compound in exactly the wrong direction. A student practises, hits 38 once, remembers 38, and walks into an exam where their real number is closer to 30 and drops further under pressure. They were not careless. They were calibrated on the wrong figure.
The practical correction is unglamorous: take several runs, use the median rather than the best, and practise at least sometimes under conditions that resemble the test — cold, once, no restarts. A protocol for getting a reading that means something takes about ten minutes.
But won't AI handle all this?
This is the first question anyone asks now, and it deserves a fair answer rather than a defensive one.
The honest concession first: less prose is typed from scratch than five years ago. AI drafts emails, summarises documents and produces first versions of things people used to write themselves. Voice input has genuinely improved and is genuinely useful in some situations.
But none of that removes the need to interact quickly and accurately with a digital system. People still search, edit, correct, enter data, write instructions and communicate privately. If anything there is more editing than there used to be, because reviewing and fixing machine-generated text is itself keyboard work, and there is a great deal more machine-generated text.
Voice also has practical limits that rarely appear in demonstrations. It struggles with proper nouns, with structured data, with anything needing exact formatting. It is impractical in a shared office, at a bank counter, in a classroom, or anywhere a conversation is already happening. And a great deal of work is confidential enough that speaking it aloud is not an option.
My reading is that typing has become less likely to be the bottleneck in producing a first draft, and more likely to be the bottleneck in everything that happens after it.
AI changed what we type. It did not change whether we need to.
What can actually be done
None of this requires a policy change to begin, which is the useful part. A few things that would move the number:
Measure before assuming
Most institutions have no idea what their students or staff actually type at. A single ten-minute exercise across a classroom or an office produces a distribution, and the distribution is usually a surprise. You cannot argue for time on a timetable without it.
Fifteen minutes daily, not a term of theory
Typing is a motor skill and responds to distributed practice far better than to concentrated instruction. Short daily exposure beats a computer-lab period once a week. It also fits into existing schedules without displacing anything.
Accuracy before speed
A learner who chases speed first grooves errors and then has to unlearn them. Speed arrives as a by-product of not hesitating and not correcting, so the accuracy work is the speed work. The four stages of learning to type covers the sequence, including the fortnight where doing it properly makes you slower and most people quit.
Rehearse the format, not just the skill
If the goal is an exam, the exam's scoring rules matter as much as raw speed — error penalties, minimum duration, whether corrections are allowed. Practising in a comfortable format and testing in a strict one is how prepared candidates still fail.
- Policy names coding and digital literacy; typing is assumed and largely untaught.
- 82.4% of 4,830 graded tests here fall below the 35 WPM SSC CHSL requires — and 75.8% of typists miss it at their personal best.
- This is a self-selected, motivated population. The general figure will not be better.
- Two in three typists are slower when the result counts, and people quote a best run that sits ~18% above their typical one.
- The gap is eight to ten words a minute — months of deliberate practice, not years.
I did not set out to write about education policy. I set out to build something because a queue at a bank counter was moving slowly for a reason nobody was naming. The data turned out to say something larger than the observation did, which is the useful thing about collecting it.
If you want to know where you personally sit against that 35 word line, it takes about a minute to find out. Most people are surprised, and usually not in the direction they expect.
Frequently asked
What typing speed does SSC CHSL require?
Is typing still a relevant skill in the age of AI?
What is a good typing speed for a job?
Why do students type slower in an exam than in practice?
How much practice does it take to reach 35 WPM?
A note on the data. All figures come from TypeLords activity to 24 July 2026: 4,830 graded typing tests by 980 typists, with attempts flagged by our anti-cheat system excluded. Around 94% of platform activity originates in India. This is a self-selected population of people who chose to use a typing platform, and should not be read as representative of the general public. Exam thresholds are from SSC notifications and change between years — confirm against the official notification for your exam.