Guide

Creator Voice Guide Template: Every Field, Ready to Copy

The document that lets several chatters write as one person: register, length, verbal tics, forbidden subjects and three sample messages, ready to fill in.

Published 28 July 2026

9fields in the voice guide template below, ready to fill instructure of this guide
3sample messages the creator writes herself, at the bottom of the documentstructure of this guide
1.5Mmessages analysed, where ending a sales message on an ellipsis measured as the worst closerour data
42,852conversations behind the prohibitions in the sales-message fieldour data

An agency inherits a problem the day the second person starts writing in the same inbox: the fan is talking to one woman, and two people are answering. Vocabulary gives it away first: “babe” from one writer, “hun” from the next, and the person paying registers the seam inside a week.

A voice guide is the document that closes that gap. This page is the document itself: nine fields and three real messages. Copy it, fill it, keep it to two pages.

What is a voice guide template for, and who uses it?

It answers one question (how does she sound) and deliberately answers nothing else. What to do next belongs to a chatting script; what this fan was told last night belongs to a handover.

  • The creator fills the identity, the forbidden subjects and the three sample messages. Nobody else can.
  • Every chatter on the account reads it before their first shift and consults the verbal tics field mid-thread.
  • Whoever reviews messages uses it as the reference for what counts as drift, which turns a vague “that does not sound like her” into a specific line.
  • Whoever hires uses the sample messages as the test piece. Give a candidate the guide and a thread, and see what comes back.

One creator, one document, one owner. If it covers two accounts it is a house style, and a house style is the thing you were trying to avoid.

What does the complete voice guide look like?

Nine fields. Copy the header block and the table, keep the middle column as guidance, fill the right-hand column.

Creator: ___________
Document owner (one named person): ___________
Written: __________      Last reviewed: __________
Field What goes in it Yours
Identity Name used, age shown, city, relationship status, job or studies. Mark what is true and what is played ___
Register How she speaks to a stranger versus a regular. Warm or dry, teasing or direct, formal or not ___
Message length Typical length, hard maximum, and whether one thought equals one message or several ___
Punctuation and formatting Capitals, full stops, exclamation marks, emoji use, abbreviations, typos left in or cleaned up ___
Verbal tics Ten words and phrases she uses constantly. Ten she never uses. Greetings and sign-offs ___
Rhythm How fast she answers, whether she writes in bursts, when she goes quiet ___
Fixed facts and stories Three or four stories she reuses, with their details locked: the dog’s breed, the job, the hometown ___
Forbidden subjects What she does not discuss, what she never promises, where she ends a conversation ___
Sales-message rules How the offer is written and closed. Voice yields to what works here ___

Fill them top to bottom, with one exception. Identity constrains register, register constrains length, but verbal tics is the field a new chatter rereads most, so leave it until you have read enough of her outbox to know what actually repeats.

What do the three sample messages look like?

Three slots (a first reply to a new subscriber, a warm mid-thread message with nothing being sold, and a refusal), each pasted from her outbox, unedited. They sit at the bottom of the document and they are worth more than everything above them.

SAMPLE 1: first reply to a new subscriber
Written by her, not by us:
_________________________________________________

SAMPLE 2: mid-thread, warm, nothing being sold
Written by her, not by us:
_________________________________________________

SAMPLE 3: a refusal, or a request she will not do
Written by her, not by us:
_________________________________________________

Rule: these are pasted from her real outbox, unedited. If a manager
wrote them, delete them. They describe an imagined voice.

The refusal sample is the one people skip and the one that saves the most trouble. Every chatter will be asked for something the creator does not do, usually late in a shift, and without a model they will each invent their own way of saying no. Some of those ways lose the subscription.

How do you fill in each field?

From her outbox, not from a questionnaire. Read a few hundred of her sent messages and mark every habit that repeats. That record is the guide; what she tells you about her style is not.

  • Identity. Write the line between true and played explicitly. A detail nobody agreed on gets improvised differently by two chatters, and the fan is the one keeping score.
  • Register. Describe it against something. “Warmer than a colleague, cooler than a girlfriend” is usable at 2am; “authentic and engaging” is not.
  • Length. Count real messages rather than estimating. Most creators write shorter than their team writes for them, and long messages are the most common single tell.
  • Verbal tics. The banned list matters more than the approved one. One word she would never type undoes a page of correct register.
  • Rhythm. Include the silences. A reply that lands within seconds at 4am, every night, stops reading like a person.
  • Fixed facts. Lock the details, not the story. A dog that changes breed between shifts is the classic tell, and it is always a small detail, never a big one.
  • Forbidden subjects. Written by her, in her words, and phrased as what she will not do rather than what she prefers.
  • Sales-message rules. Three lines, and they override the voice: close on a statement, give the figure once and plainly, and let her register take over again in the next line. This is the field that stops a soft, trailing-off register from producing sales messages that end in an ellipsis (the worst-performing closer in our data) and stops a considerate register from closing on a yes/no question, which costs several points of conversion. Add the third prohibition explicitly: no personal callback at the exact moment of the pitch, however natural it feels.

Which three mistakes make a voice guide useless?

The same three every time, and none of them is about the writing.

  1. Adjectives instead of examples. “Playful but classy” gives a new chatter nothing to copy. Three of her real sentences give them everything. If you had to delete either the description or the samples, delete the description.
  2. Written by the manager. A guide assembled from what the creator says about her style describes a voice that has never sent a message. Build it from the outbox, then let her correct it. That order, never the reverse.
  3. Too long to open. Past two pages the document becomes onboarding material: read once, never consulted, and drift resumes in week two. Every field you add costs you the fields already there.

A fourth, quieter failure: writing it once and never touching it again. A quality control pass that spots a tone drift is worth one added line, not a rewritten section.

When do you revise it?

On a fixed date, and on triggers. Never during a shift, because a voice changed mid-shift is a voice that exists in two versions for a day.

  • On a schedule, monthly or quarterly, with the person who owns the document named in it.
  • When a new chatter joins. Their first week produces the best list of questions the guide fails to answer. Write the answers in, and see hiring chatters for how to test against it before they start.
  • When a drift is deliberate. A chatter who writes something better than the document either changes the document or stops. Undecided drift is how a team ends up with three voices and one guide.
  • When the creator’s own life changes. A moved city or a new job breaks the fixed facts in every thread at once.
  • When the shift rotation gets finer. More writers on the same inbox means more drift per week, and the document has to get more specific to hold.

What it will never fix: the memory of an individual conversation. A voice guide keeps everyone sounding like one person. Keeping them saying the same things is a different document.

Frequently asked questions

Who should write the voice guide, the creator or the manager?

The manager assembles it, the creator supplies the raw material. A creator asked to describe her own writing describes an intention; her sent folder shows the habit, and the habit is what a chatter has to reproduce. The sample messages must be hers word for word, and the forbidden subjects field must be hers too, because nobody else can decide what she will not say.

How long should the document be?

Two pages, and treat that as a hard limit rather than a target. A voice guide is consulted mid-shift by someone with an open inbox, so anything that does not fit on a couple of screens is a manual, and manuals do not get opened. Every field below earns its place; nothing else does.

Can one voice guide cover several creators?

No. The template travels, the content never does. A shared guide produces a house voice, which is exactly the thing fans notice when they compare messages between two accounts run by the same agency. The stage map of a conversation can be shared; the way a person sounds cannot.

What if a chatter naturally writes better than the guide says?

Then the guide is out of date, not the chatter, but you change the document before you change the messages, otherwise you have two voices again. Log the drift, review it at the next revision, and either adopt it as the voice or correct it. What you cannot do is leave it undecided.

Does a voice guide stop chatters contradicting each other?

It stops them sounding like different people. It does not stop them contradicting facts, because the guide has no memory of what this fan was told last night. That is what a handover carries. Give a team a flawless guide and no handover, and every message will sound like her while half of them deny what she said yesterday.

How do you check the guide is actually being followed?

Read real messages against it, in short passes, regularly. Pick a handful of threads per chatter, mark every line that could not have come from the document, and add one line to the guide for each drift you decide to allow. A guide that has not changed in months is usually a guide nobody consults.

See what it looks like in practice

The justonedash chatbot holds the conversations, keeps each creator’s voice and works around the clock.