The first-draft problem

Most writing advice eventually lands on the same insight: your first draft should sound the way you talk. Vivid. Concrete. Specific. Sentence rhythm that breathes. Strunk and White, Stephen King, every editor you've ever had — same note, different decade.

And yet when you sit down at a keyboard with a blank document, you almost never write that way on the first pass. You write more cautious, more formal, more "this could be in a newsletter". You write paragraphs you wouldn't say to anyone, in a register you'd never use out loud.

That gap between how you write a first draft and how you talk is where most of the "this sounds AI-generated" complaint comes from. It's not the AI — it's the medium. Anyone typing into a blank page becomes a slightly more cautious version of themselves.

What dictation actually changes

When you dictate the same idea out loud, three things shift, in this order:

1. You're more committed to your sentences

Typing has a backspace key. Speech doesn't, or at least the friction of restarting a sentence out loud is higher than backspacing one. So you finish sentences instead of fragmenting them. You commit. The dictated paragraph has fewer abandoned thoughts trailing off into "and yeah".

2. You sound like a human explaining something

Read this typed version of an explanation:

The product enables users to leverage voice input across all applications in the Windows ecosystem.

Now the spoken version:

If you hold the hotkey and talk, the text just shows up wherever your cursor is. Doesn't matter what app you're in.

Same information. The second one is what someone would actually say. The first one is what they'd write when they're being careful and have a blank document making them nervous.

The dictated version is the better one for almost every audience. It's clearer, it's shorter, and it sounds like an adult talking to another adult.

3. You stop ranking the importance of each word

When you type, every word costs a keystroke. Subconsciously you start optimizing — leave out the connecting word, cut the example, skip the qualifier. The result reads dense and pruned, in the way that legal contracts read dense and pruned.

When you speak, words are free. So you include the connector ("it's worth saying that..."), you include the example ("for instance, if you're..."), you include the qualifier ("at least in my experience"). And that's the thing readers latch onto. The connective tissue is what makes prose feel human.

The second-draft trick

This is the workflow you actually want, once you've done it a few times: speak the first draft, edit it typed.

The speaking pass gets you raw, voice-y prose with everything in it. The typed editing pass tightens it, fixes the rambling bits, structures it. You end up with something that reads like writing but sounds like talking — which is what every reader unconsciously wants.

The reverse order doesn't work. If you type the first draft, you can't easily talk it into being warmer afterwards — the cautious version is already in your head.

Try it on something low-stakes first. A LinkedIn post, a "what we learned" memo for your team, a longer Slack reply you'd normally over-format. Dictate the whole thing in one take without correcting yourself, then read it back and clean up the messes. You'll notice the voice is more present than it usually is. That's the point.

Where this matters most

Three audiences for whom this is a multiplier rather than a marginal gain:

Founders writing their own LinkedIn

The dead giveaway of a ghostwritten LinkedIn post is over-structure. Three crisp paragraphs, each ending on a punchline, the whole thing too clean. Founders who dictate post in a rougher, more honest register that performs measurably better in the same feed. The platform's own engagement signals reward it.

Newsletter writers

Open rate is mostly subject line. Read-through is mostly voice. People who sound like a real person in paragraph one keep readers around for paragraph six. Dictating your first draft is the cheapest, fastest way to keep the voice in.

Product marketers

The features page that reads "leverage cutting-edge AI to streamline your workflow" was typed. The one that reads "press a key, talk, the text shows up wherever your cursor is" was either dictated or written by someone who imagined themselves dictating. Readers can tell which is which within ten seconds, and it determines whether they buy.

The verification step

The honest counter-argument: dictation is messy. You will mishear yourself. The model will mishear you back. Names and brand terms will get spelled phonetically the first time. Sentences will trail off in places. None of this is a deal-breaker, but it means dictation produces a draft, not a final product.

So the speed advantage we wrote about in the previous post is real but smaller than the headline. You always re-read. You always fix three or four words. Net-net dictation is still faster than typing for any prose longer than a sentence, just not the 3x advertised.

But the voice gain — the bit where your prose sounds like you — that one doesn't get smaller after editing. The voice is in the draft from the moment you opened your mouth. Editing keeps it there. That's the win, and it's the win that compounds across every paragraph you write for the next ten years.