Stop typing at your AI. Start talking to it.

The short version: For the last few months I've spoken about 95% of what I put into AI tools rather than typing it. The output got noticeably better, not because the models changed, but because speaking forces me to give them far more context than typing ever did. If you want sharper answers from AI, the fastest lever isn't a cleverer prompt. It's talking to it like you'd brief a smart colleague.

Key takeaways

• Typing makes you compress and edit yourself before the words land, so you hand AI a thin brief and get a thin answer back.

• Speaking removes that friction. You ramble, backtrack, add context, and all of it gives the model more to work with.

• The research is clear on speed: conversational speech runs around 150 words per minute against roughly 40 for average typing, and a well-known Stanford study found speech input about three times faster than typing on a phone.

• The bigger gain isn't speed. It's that voice quietly makes you a more thorough, more honest briefer.

• The practical move: talk to AI the way you'd talk through a problem with a colleague, not the way you'd fill in a form.

Typing makes you a worse explainer

Here's the thing nobody tells you about prompting. When you type, you edit yourself before the words ever reach the screen.

You compress. You leave out the bit of context that felt like too much effort to write down. You hand the model a neat, summarised version of the problem rather than the actual problem. And then you wonder why the answer comes back generic.

It isn't the model being shallow. It's you giving it a shallow brief, because typing taxes you into being brief.

When you speak, that tax disappears. You ramble. You go off on a tangent and come back. You say "actually, the real issue here is..." three sentences in, which is usually the most useful thing you say. All of that is context. The richer and messier the input, the more the model has to work with, and the sharper the output gets.

The research backs the speed, and the speed is the easy part

There's solid evidence on the mechanics. Conversational speech runs at around 150 words per minute, against roughly 40 for the average typist. A well-known Stanford study found speech input was about three times faster than typing on a phone, in both English and Mandarin.

So the speed gap is real and well documented. But speed isn't what changed my results. Speed just gets the words down faster.

The part the studies don't measure, and the part I'd point to, is what the friction of typing does to the quality of your input long before speed enters the picture. A keyboard makes thoroughness feel expensive, so we ration it. A microphone makes thoroughness free, so we stop rationing. That's the real shift.

We've quietly been here before

This reminded me of something about how work used to be organised.

There used to be a clean split between having the idea and getting it down on the page. You could think something through out loud and let someone or something else handle the mechanics of capturing it. The thinking and the transcribing were two different jobs.

Then computers arrived on every desk and we quietly collapsed that split. Everyone became their own scribe. We simply accepted that the person with the thinking to do should also be the one mechanically formatting it, and we've worked that way for so long it stopped feeling like a choice.

Voice plus AI hands the split back. You think out loud. Something else captures it. Except this time the thing doing the capturing also understands what you meant, and can act on it.

That's the modern version of the old arrangement, and it's a meaningfully better one.

Why this matters beyond a workflow tweak

It's tempting to file this under productivity tips. Talk instead of type, save a few minutes, move on.

I think it's bigger than that. How you get information into these tools is becoming as important as which tool you pick. The people getting the most out of AI aren't the ones with the cleverest prompts. They're the ones willing to give it the full, unfiltered context of what they're actually trying to do, the way you would when you're thinking out loud with a colleague who's genuinely trying to help.

Which, when I think about it, is most of my job anyway. Taking something complex and tangled, talking through it until it's clear, then turning that clarity into something that actually moves. The tools have just caught up to the way good thinking already works.

So talk to it. Properly. Like you'd brief a smart colleague, not like you're filling in a form.

You'll feel slightly mad doing it. The results are worth it.

How to get started

If you want to try it, the setup is simple. You need a voice dictation tool that types into whatever you're already using. I use Wispr Flow, which runs across every app, so I can speak into Claude, ChatGPT, email, or notes without switching anything. Then change one habit: instead of writing a tight prompt, talk through the whole problem out loud. Context, background, what you've already tried, what good looks like. Let it be messy. The mess is the point.

FAQ

Is talking to AI actually better than typing, or just faster?

Both, but for different reasons. Conversational speech runs around 150 words per minute against roughly 40 for the average typist, so speaking is meaningfully faster for most people. The less obvious benefit is quality: because speaking is low-effort, you naturally give the AI far more context than you would when typing, and more context produces sharper output.

What do I need to dictate to AI?

A voice dictation tool that converts speech to text inside the apps you already use. I use Wispr Flow, which works system-wide, so you can speak into any AI chat tool, your email, or your notes. Your phone and laptop also have built-in dictation if you want to test the idea before adding anything.

Won't the AI get confused if I ramble?

The opposite, usually. Rambling includes the context, caveats, and half-formed thoughts that a tidy typed prompt strips out. Modern AI tools are good at working out what matters from a longer, messier input. A bit of structure helps, but you don't need to script it.

Does this work for serious or professional work, or just quick tasks?

It's most valuable for complex, ambiguous work, where explaining the full situation matters. Strategy questions, drafting, thinking through a problem. The harder the thing you're working on, the more the extra context pays off.

Do I look strange talking to my computer?

A bit. You get used to it, and so does everyone around you. The quality of the output makes it an easy trade.