guides
How to Write a Chatter Style Guide for a Model's Voice and Tone
A fan messages three chatters in one week and gets three different personalities. Here's how to write a style guide that keeps a model's voice consistent no matter who's typing.
A top-spending fan messages his favorite model on Monday. She's playful, uses lowercase, drops three emojis a message. On Wednesday, a different chatter is covering the account. Suddenly she's formal, no emojis, calls him "hun" out of nowhere. The fan notices. He doesn't say anything — he just spends less. Then he stops replying.
That's not a chatter problem. That's a documentation problem. If the model's voice only lives in one chatter's head, you don't have a brand — you have a person, and people are inconsistent, get sick, quit, and get replaced.
What should a chatter style guide actually include?
A working style guide needs at minimum 6 elements: tone descriptors, vocabulary rules, emoji/punctuation habits, message length norms, banned phrases, and 10-15 real sample messages pulled from the model's best-performing conversations. Anything shorter than that leaves too much room for a new chatter to guess — and guessing is exactly what creates the inconsistency fans pick up on.
Think of it less like a brand book and more like a script bible. It's not there to make the model sound robotic — it's there so five different chatters sound like the same person.
Start With Tone Descriptors, Not Adjectives
"Friendly and fun" tells a chatter nothing useful. Every model claims to be friendly and fun. You need descriptors specific enough that two different people would type the same kind of message from them.
Pick 3-4 tone words that are specific and slightly contradictory on purpose — e.g., "playful but a little bratty," "warm but never needy," "confident, teases more than she compliments."
Attach a rule to each word. "Bratty" means she pushes back on requests before agreeing. "Never needy" means she never asks "are you still there?" more than once.
Define energy level by time of day. Late-night messages run flirtier and shorter. Morning messages run warmer and a bit more check-in style.
Lock Down Vocabulary and Formatting Rules
This is the part most agencies skip, and it's the part fans notice fastest. Word choice and formatting are the fingerprint of a voice — more than tone, honestly.
List approved pet names in order of familiarity (new fan vs. VIP fan) so a chatter doesn't call a day-one subscriber "baby" the way she talks to a $500/month regular.
Set capitalization rules. All lowercase? Sentence case? Never both — pick one and put it in writing.
Cap emoji usage. "1-2 emojis per message, never more than 3 in a row" is a rule a new hire can follow on day one. "Use emojis naturally" is not.
Define message length. If she texts in short bursts of 4-8 words, write that down — chatters coming from other accounts often default to longer, more "customer service" style replies.
Note punctuation quirks. Does she use "..." a lot? Never uses periods at the end of a text? These tiny habits are what make a voice feel like a real person instead of a script.
Build a Banned List — It Matters as Much as the Style Rules
Every style guide needs a section of things the chatter should never say, because new hires default to generic sales language under pressure. Ban phrases like "as an AI," obvious upsell language ("check out my exclusive content!"), overly formal greetings ("Hello, how may I assist you today?"), and anything that sounds like a template. If a fan can tell he's talking to "a chatter" instead of "her," the illusion breaks and so does the sale.
Give Chatters Real Sample Conversations, Not Just Rules
Rules describe a voice. Examples teach it. Pull 10-15 real message exchanges from the model's highest-converting chats — a mix of casual chat, a PPV pitch, a re-engagement message to a fan who went quiet, and a response to a difficult or demanding fan. New chatters should read these before they ever send a live message, not after.
This matters even more if you're onboarding your first hire rather than your fifth — our guide on how to hire and train your first chatter as a solo creator walks through the ramp-up process alongside the style guide itself.
Keep the Guide Where Chatters Actually Work
A style guide in a Google Doc nobody opens mid-shift doesn't help anyone at 11pm when a chatter is juggling four conversations. The guide needs to live where the work happens — ideally pinned inside the same workspace chatters use to message fans.
This is also where a shared inbox setup pays off. CRMChat is a Telegram-native CRM built for agencies running multiple model accounts and multiple chatters, and it includes a separate chatter role with limited permissions — so you can onboard new chatters into a shared workspace without handing over full account access, while every chatter can still see the same pinned style notes and past conversation history for context before they type a single reply.
If you're running an AI layer alongside human chatters, consistency gets even more important. CRMChat's AI response automation uses a knowledge base you provide to keep the model's voice consistent even when the AI is handling replies overnight — meaning your style guide isn't just for humans, it's the same document the AI reads from.
Review and Update the Guide Every Few Weeks
A style guide isn't a one-time document — voices drift as fans respond to certain phrases more than others. Set a recurring 15-minute review with your top chatter:
Pull the top 5 highest-converting messages from the past two weeks and check if they match the current guide.
Add any new phrases or jokes that became "canon" organically and are working.
Remove anything stale — inside jokes tied to an old promo, references fans stopped reacting to.
Ask every active chatter if any rule feels unnatural to follow — a rule chatters keep breaking is usually a rule that needs rewriting, not enforcing harder.
The goal isn't a rigid script. It's a voice consistent enough that a fan spending $500 a month never has a Tuesday where "she" suddenly feels like a stranger — because that's the moment he starts wondering if he's actually talking to a person at all.
