Dictate your prompts to Claude (and ChatGPT): voice in the service of AI

More and more people spend their day talking to an AI: asking Claude a question, having ChatGPT rephrase an email, requesting an outline, some code, a summary. And everyone ends up noticing the same thing: the answer is only as good as the request. A sloppy prompt gives a sloppy answer. And on a keyboard, we almost always end up being sloppy. This is exactly where voice dictation changes the game, and not just to save time.

The answer is only as good as the prompt

This is not just an opinion, it is what the AI companies themselves recommend. In its prompt engineering guide, Anthropic keeps making the same point: a good prompt names the task explicitly, gives the context, says who the result is for, the expected format, and what a successful piece of work looks like. "Write something about marketing" goes nowhere; "Write an 80-word follow-up email, warm in tone, for a client who hasn't replied in two weeks" gives you a usable result on the first try.

In other words, the value lives in the details you provide. The more complete and structured your request, the better the answer. The problem is that providing those details on a keyboard is tedious, so we skip them.

On a keyboard, we write cut-rate prompts

Type, and you will catch yourself keeping it as short as possible. We write "summarize this", "fix this", "give me an outline", without the context that would make all the difference. Why? Because typing is slow and tiring: we abbreviate, we skip the specifics, we stop at the first phrasing that does the job. On a smartphone it is even worse: composing a long prompt with your thumbs on a touchscreen is sheer misery, so we cut things down even further.

The result: we under-specify our request, the AI fills in the gaps as best it can, and we run through one back-and-forth after another to recover what we could have said from the start.

Voice: faster, and above all more complete

Speaking removes that friction. We naturally dictate around 200 words per minute, roughly four times faster than typing. But speed isn't even the main point. The main point is that speaking is our most natural way of expressing a thought: you state your request as if you were briefing a colleague, with the context, the nuances and the constraints that go with it, without the effort that pushes you to cut corners.

When we dictate, we spontaneously say what we wouldn't bother to type. Compare:

The second version took no more effort: it was simply spoken instead of typed. And it is the one that gives you a genuinely usable answer.

How to dictate a good prompt

No complicated method required. As you dictate, keep in mind the elements a good instruction contains, and say them in the order they come to you:

• The role you want the AI to play, if it helps ("as a lawyer", "like a senior developer").
• The context: who the result is for, in what setting, with what starting information.
• The precise task: exactly what you want, not a vague intention.
• The expected format: length, structure, list or paragraph, tone.
• The constraints: what to keep, avoid, or double-check.

Just think out loud, as if you were explaining your need to someone sitting across from you. It is the fastest way to get a rich prompt, and it is exactly what the keyboard discourages.

Claude, ChatGPT, and all the rest

The benefit isn't tied to any one assistant. Whether you use Claude, ChatGPT, Gemini or another model, you type your requests into a text field. Voice dictation that works at the system level types into any field, so straight into your AI's input box, on the web as well as in a desktop or mobile app. The same reflex serves you when you dictate to a coding assistant: you describe what you want out loud, in full.

The detail that makes the difference: a clean prompt, not a raw stream

Dictation has a well-known pitfall: spontaneous speech is full of hesitations, "ums", repetitions and false starts. A raw transcript leaves them all in, and you end up with an off-putting block of text to paste into Claude. This is where a dictation tool with AI cleanup really matters: it turns your stream of speech into clean, punctuated, structured text, ready to serve as a prompt. That is exactly what Fast Dictate does: you speak, a clear prompt appears in the field, with a single shortcut, in any app on Windows and Mac. You keep the naturalness of voice without the mess.

Frequently asked questions