Long-form dictation: speak the words, build the structure after

Last updated 11 September 2026

Why long dictations fall apart

Short dictations work because the whole thing fits in your head. A three-sentence reply has no structure to lose.

At two thousand words that stops being true, and speech gives you none of the help typing does. You cannot see the page. You cannot scroll back. You cannot look at the heading three sections down and remember what this part is supposed to do. So you drift, repeat yourself, and arrive at the end having covered the same ground twice and skipped something.

The fix is not to speak more carefully. It is to do the structural work in a different mode, before you start talking.

The method

1. Outline by keyboard, in headings Six to twelve headings in the document itself. Not a plan on paper — headings you will dictate between. This is the step people skip and it is the whole method.
2. Add one line under each heading What this section has to do. "Explain why the obvious approach fails." Typed, not spoken. Now every section is a short dictation with a brief.
3. Dictate one section at a time Two hundred to four hundred words. Latch the microphone; do not hold a key. Stop at the end of the section even if you are flowing.
4. Do not look at the screen while speaking The same rule as everywhere. Watching text appear pulls you into editing mid-thought.
5. Fix only recognition errors between sections Not the prose. A quick pass for wrong names, nothing else.
6. Restructure by keyboard, on a different day if you can Moving, cutting, joining, ordering. This is where the document becomes good, and it is a keyboard activity.

Why sections rather than one long pass

That last row is the useful one. A section with a one-line brief is a question, and answering a question aloud is something people are good at. "Write a chapter" is not a question.

What to expect from a dictated draft

It will be longer than a typed one, looser, more repetitive, and more conversational. Some of that is a problem and some is an improvement.

• Expect 20–40% more words than you need. Budget a cutting pass. Cutting is easier than expanding, which is the trade.
• Expect repetition. You will make the same point twice in different words. Keep the better one.
• Expect better rhythm. Spoken sentences vary in length; typed ones drift towards uniformity. This is a real gain.
• Expect weaker transitions. Speech connects with "and" and "so". Written prose needs more, and that is a revision job.
• Expect the good bit to be late. The most useful paragraph is often in the last quarter of a section, after you stopped performing.

The things that ruin it

1. No outline. The single cause of most bad long dictations.
2. Editing while speaking. Look away from the screen.
3. Untaught vocabulary. In long-form work the same fifteen proper nouns appear on every page. Untaught, you will spend the revision pass fixing the same name forty times.
4. Sessions that are too long. Forty minutes of dictating is exhausting in a way typing is not, and quality falls off sharply. Twenty minutes, then stop.
5. Trying to revise by voice. Revision is precision work. Keyboard.