Does Typing Speed Still Matter With AI Autocomplete?

The Short Answer

> Quick answer: AI autocomplete removed a lot of boilerplate typing, and it added a huge amount of prompt writing, review and correction — which is all typing. Input speed matters less for producing code and more for thinking out loud, editing and communicating. Touch typing isn't obsolete, but the reason to learn it has shifted from throughput to friction.

The "typing is dead" argument shows up every few months and it always sounds plausible for about a paragraph.

The version I hear most: an assistant writes the function, so why practise typing it? Fair question. My honest experience is that I type more now than I did three years ago, just different things.

So rather than argue the abstract, here's what actually changed in the day-to-day, and where the skill still earns its keep.

What AI Tools Genuinely Changed

Boilerplate is gone, and good riddance. Nobody's hand-typing a constructor, a test scaffold or a config block anymore, and those were a meaningful share of daily keystrokes.

Repetitive edits went too. Renaming across files, filling in switch cases, converting a data shape — all things that used to be typing and are now a request.

What replaced them is prose. Prompts, follow-up corrections, review comments, commit messages, design notes. The centre of gravity moved from typing code to typing English about code, and English is where most people are actually faster.

The net effect on my own keystroke count was close to flat. What changed was the mix, and the mix now favours sustained prose over precise symbol entry — which is a friendlier skill profile for anyone who touch types.

Where Input Speed Still Decides Things

Thinking out loud is the big one. When you're working through a problem in a scratch file or a chat window, slow typing genuinely interrupts the thought. There's a real cognitive cost to your hands lagging your head, and it's the same reason people who can't touch type find long-form writing harder.

Interviews are the sharp case. Live coding while someone watches, with no assistant, at whatever speed your fingers manage. That's still a common format and fumbling looks worse than it is.

Meetings and pairing sessions too. Typing notes while listening only works if typing is automatic — the moment it takes attention, you stop listening properly.

And correcting AI output is typing. Not much of it, but precise and symbol-heavy, which is the exact weak spot most people have. The tasks got shorter and more fiddly, which arguably raises the value of accuracy over raw speed.

The Accuracy Argument Got Stronger

Here's the shift I didn't expect. Speed matters less than it did and accuracy matters more.

When you were typing a hundred lines, a typo was one of a hundred and you'd catch it. When you're typing a twelve-word prompt or a two-character fix, a typo is a much larger fraction of the input and it changes the output.

Mistyped prompts are the modern version of a compile error. You get a confidently wrong answer to a question you didn't ask, and then spend two minutes working out why.

So if I were choosing one thing to practise now, it'd be accuracy rather than speed — and the two are related in ways people underestimate, as our piece on why accuracy drops as speed increases goes into. Slowing down 10% to gain 3% accuracy is a better trade in 2026 than it was in 2020.

The Jobs Where Nothing Changed at All

The whole debate is unusually developer-centric, and most typing in the world isn't developers typing code.

Transcription is unchanged. Medical and legal transcription still requires a human matching audio at speed, and while automated transcription got good, the correction pass is typing and the accuracy bar is unforgiving.

Data entry is unchanged. Reading a form and keying it in is still the job for a very large number of people, and throughput targets are measured in keystrokes per hour rather than in prompts. Our note on WPM to KPH covers how those roles actually measure you.

Customer support is unchanged, and arguably got more typing-heavy as chat replaced phone queues. Support agents type all day, under time pressure, while reading.

Students are unchanged too — essays, notes and exams are still typed, and typing speed correlates with how much of a lecture you capture. So when someone announces that typing is obsolete, it's worth asking whose job they're describing.

What the Data Suggests

Developer surveys have tracked AI tool adoption climbing steadily, with the large majority of professional developers now using assistants regularly — the annual Stack Overflow developer survey is the usual reference point for those figures.

What those surveys consistently don't show is developers reporting less keyboard time. The reported change is in what the time is spent on, with review and verification taking a larger share.

I'd treat any specific productivity claim in this area sceptically, in both directions. The measurement is hard and the incentives to overstate are enormous.

What's uncontroversial is narrower and more useful: the assistant produces a draft, a human reviews and corrects it, and the review step involves reading and typing. Nobody's proposing a workflow where the human stops using a keyboard, and until someone does, the skill retains its value.

Voice Input Didn't Replace It Either

The other periodic claim is that dictation makes typing redundant. Voice recognition genuinely got good, and it still hasn't displaced the keyboard for most work.

The reasons are practical rather than technical. Open offices make dictation socially awkward. Code is nearly unspeakable — try dictating a regex. And editing by voice is far slower than editing by keyboard, so any task involving revision favours hands.

Dictation is excellent for first drafts of prose, and I use it for exactly that. It's poor at everything after the first draft, which is where most of the work happens.

We compared the two properly in voice typing versus keyboard. The short version: they're complementary, and the keyboard wins on precision by a distance that hasn't narrowed much.

The Skill That Quietly Got More Valuable

One thing genuinely became more useful, and it isn't speed. It's the ability to write a clear, complete instruction on the first attempt.

That's a writing skill sitting on top of a typing skill. If typing costs you attention, your prompts get shorter and vaguer, because writing the longer version feels like work. Then you get worse output, and you blame the model.

I notice this in myself. On days when I'm tired, my prompts get terse and my results get worse, and the cause is friction rather than thinking.

Someone who touch types writes the three-sentence version without noticing they've done it. That's a small advantage repeated fifty times a day, and it compounds in a way that's hard to see from the outside.

Which lands somewhere unexpected: the argument for typing practice in 2026 is stronger for writing than for coding. Fluent hands mean you express the whole thought instead of the abbreviated one, and that applies to prompts, review comments, documentation and messages to colleagues equally.

What I'd Practise Now

Touch typing, still, for anyone who hasn't learned it. The argument for it was never really throughput — it's that not looking at your hands frees attention for thinking, and that hasn't changed at all.

Accuracy over speed, for the reasons above. A 70 WPM typist at 99% accuracy is more effective than a 95 WPM typist at 94%, and the gap widens as inputs get shorter.

Punctuation and the number row, because that's what correcting and prompting involves. General prose practice won't touch either.

And not much time on it. Fifteen minutes a day for a couple of months, then maintain with a weekly test and stop thinking about it. The right amount of effort for this skill in 2026 is modest and non-zero — which is a boring conclusion, and I think it's the correct one.