Also called: how many samples, corpus size
Writing sample size is the amount of an author's own text a system needs before per-author style measurements stabilise, and the binding constraint is usually variety of contexts rather than raw word count.
Coverage of the registers you actually write in matters more than total words.
Word count is the number people ask about because it is easy to check. It is the wrong constraint to optimise alone. Ten thousand words of one document type produce a confident profile of one register and a misleading profile of you.
The useful framing is coverage. List the media you write in, the audiences you write to, and the languages you write in. A profile is ready for a cell in that grid when it has seen that cell, and it is speculating about every cell it has not seen.
| Coverage | What the profile can do | What it cannot do |
|---|---|---|
| One medium, one audience, one language | Reproduce your voice in that cell | Predict your register in any other cell |
| Several media, one audience | Handle format shifts within that relationship | Handle a change of reader |
| Several audiences and media | Select a register per task | Cover a language it has not seen |
| More than one language | Preserve language-specific register habits | Substitute for samples in a third language |
Cite this page
MyWritingTwin. “Writing sample size for an AI voice profile.” Last updated 2026-09-15. https://www.mywritingtwin.com/glossary/writing-sample-size