Voice is the set of word choices, rhythms and limits making your text replies sound like one person rather than a form letter. Personality adds what the same person cares about and refuses to discuss. Written down as rules, both survive automation, left in your head, both drift within a week. The three stages run in build order.
What does a voice brief need before you switch on replies?
A voice brief needs four written decisions: who speaks, how long a reply runs, which subjects get a fixed deflection, and how AI involvement is disclosed. Finish the whole checklist in one sitting, before any agent touches a live message.
- Write a character statement of 25 words or fewer, naming age range, niche, home city and the one subject the character is always glad to discuss.
- Collect 15 messages you have already answered by hand and highlight the three sentences you would be happy to see repeated back to a stranger.
- Set a hard length rule, for example two sentences for a question and one sentence for a compliment, so no draft sprawls into an essay.
- List 10 words the character uses often and 10 the character never uses, including openers like "I hope this message finds you well".
- Name three topics with a fixed deflection line each: medical questions, anything about your home address, and anything a sponsor has not cleared.
- Store one disclosure sentence in the brief, because the FTC expects disclosures in plain language that people cannot miss (FTC guidance for influencers).
- Decide the emoji rule as a number: zero per reply, one per reply, or one only on the thread opener.
Messages addressed to a trained character need the brief in place before volume arrives. The staged approach in handling fan conversations covers approval and handover in the same build order, with disclosure written into the first reply.
When every reply sounds like a help desk ticket
Help desk tone creeps in when the notes behind an agent read like policy documents instead of speech. Rewrite the brief as spoken lines, then work the repair checklist with a week of real drafts in front of you.
- Rewrite every policy sentence in the brief as a sentence you would say out loud, and read each one aloud once before saving.
- Keep draft approval switched on in the DM agent so replies queue for your review, and leave approval on until 50 drafts pass with no edit.
- Log each draft you correct for seven days, then paste the correction into the brief the same evening, while the reason is still fresh.
- Ban three support phrases outright: "per our policy", "we apologise for any inconvenience", and "kindly".
- Require at least one contraction in every reply longer than eight words, since "I'm" and "you'll" carry more voice than "I am" and "you will".
- Give the character one recurring small detail, a 6am session or a loud neighbour, allowed in roughly one reply out of ten and never twice to the same person.
- Send 10 varied test messages from a second account, including one complaint and one compliment, and grade each answer against the length rule.
The Creator plan at $99 a month includes inbox automation and AI replies, 500 AI credits each month and one AI influencer. Voice work is iterative, so budgeting matters: every rewrite and retest of a draft consumes credits, and extra credits are sold at $1 for 10.
Drafts that answer a question nobody asked usually trace back to stale notes rather than a weak model. The walkthrough on fixing faulty DM drafts takes two common symptoms apart and shows which note to change first.
How do you keep one personality across nine platforms?
One personality survives across platforms when a single master brief governs every inbox, and per-network overrides are limited to two dimensions: reply length and formality. Creator Link Studio connects LinkedIn, Bluesky, X, Facebook, Instagram, Threads, TikTok, YouTube and Pinterest, so the same character is answering in nine places at once.
- Keep the master brief in one file and mark overrides clearly, with no network allowed a second character statement.
- Set LinkedIn replies to two sentences with no slang, X and Threads to one sentence, and Instagram to a first-name greeting plus one sentence.
- Check messaging settings on each network before connecting an agent, since account tools differ for business and creator accounts (TikTok support for creators).
- Use the same three opening greetings everywhere, so a follower who sees your replies on two networks recognises the same person.
- Pin the deflection lines across all nine inboxes at once, rather than rewriting each one per network and ending up with three versions.
- Match the written voice to the visual character, using image tools that hold one face so a playful voice is not paired with a stiff portrait.
- Score 20 sent replies against the brief on the first Monday of each month, and change the brief wherever three or more replies missed.
- Give every reviewer a seat if more than one person approves drafts, so edits land back in a single brief instead of private notes.
A brief read by two people stays honest in a way a brief read by one does not. Ask a second reader to flag any sentence the character would not say, then cut the flagged lines rather than softening them.
Monthly review is where personality either hardens or decays. Twenty scored replies take about 20 minutes, and the pattern in the misses tells you whether the problem is length, vocabulary or a missing deflection line.
