Brand voice for Zalo OA chatbots: staying consistent
Customers read your brand's messages right on their phone - the chatbot's voice is the first thing they feel. Photo: Samson Katt, Pexels.
A customer messages a shop on Zalo at 9pm. The chatbot replies with a stiff "Dear valued customer, thank you for contacting us" - while the shop's Facebook page uses a casual, friendly tone. The customer gets confused: is this even the same brand? That is what happens when a business has a chatbot but no brand voice.
Brand voice for a Zalo OA chatbot is a fixed set of rules covering how the bot addresses customers, sentence length, tone of familiarity, and a list of banned words, so the chatbot sounds like a real person from the brand rather than a script reader. With over 79.6 million monthly Zalo users (Zalo, late 2025), most Vietnamese customers already chat with businesses on this channel every day. One working session is enough to lock down three things: how the bot addresses customers, 5-7 signature words, and a banned-word list - then apply that to the chatbot script and train staff to answer manually in the same voice.
What is brand voice, and how is it different from brand tone?
Brand voice is the fixed personality a brand uses when speaking - always friendly, always concise, always addressing customers the same way. Brand tone is the shade that shifts with context - upbeat when an order is confirmed, apologetic and serious when a delivery is late. A Zalo OA chatbot needs both: a voice that never changes, and a tone that adapts to the situation. Many small shops confuse the two, so their chatbot swings between overly formal and oddly playful, and customers notice the inconsistency. For ready-made sample lines for specific hard situations, our tone of voice by situation guide includes a 5-scenario matrix with lines to use and lines to avoid.
Why does a Zalo OA chatbot need its own voice, not just correct answers?
Answering correctly is necessary but not enough to keep customers coming back. Zalo reached 79.6 million monthly active users by late 2025 and crossed 80.2 million in Q1 2026 (Zalo's own announcement), with an 83% usage rate in Vietnam - the highest among messaging apps (Decision Lab, The Connected Consumer report, Q3 2025). That means almost every one of your customers is already used to texting friends and family on Zalo in a casual, everyday voice. If your chatbot replies like a corporate call center, the contrast is obvious, and customers can feel distant or even suspect they are talking to a lifeless machine instead of a brand they trust.
On the other hand, when the chatbot's voice matches how the shop owner or staff already text in real life - a bit familiar, the right amount of emoji, a consistent way of addressing people - customers feel like they are talking to a real person from the brand, even knowing it is an automated reply outside business hours.
One 2-3 hour session with a notebook is enough to lock down a first brand voice guideline. Photo: Shixart1985, Flickr, CC BY 2.0 license.
How can a small shop define its brand voice in a single session?
You do not need to hire an agency. A 2-3 hour session with the person who understands the brand best (usually the owner) answering three questions is enough:
1. A fixed way of addressing customers: casual first-name style, polite-but-warm, or fully formal. Pick one and use it everywhere - Zalo, Facebook, website - without switching.
2. 5-7 signature words or phrases that should recur - for example, a coffee shop might choose words evoking aroma, hand-brewing, a familiar corner. These are what make the brand's voice distinct from competitors.
3. A list of banned words - overly formal phrases if the brand positions itself as youthful, slang or abbreviations if it positions itself as premium, or jargon customers will not understand.
Write all three down in one short document (half a page is enough), and that becomes the shop's first brand voice guideline - no need for anything elaborate, just clear enough that anyone writing chatbot scripts or replying manually can use it.
How do you write Zalo OA chatbot scripts that keep the right voice?
Once the brand voice guideline exists, writing chatbot scripts is simply applying the agreed formula to every sample reply. Three practical rules: short sentences, one idea per sentence - people read Zalo messages on their phones, and sentences longer than two screen lines make customers skip reading; keep the same way of addressing customers across every script - from greetings and quotes to order confirmations and apologies for being out of stock, since the most common mistake is a friendly greeting script paired with a suddenly formal apology script; and allow a small, controlled amount of natural "imperfection" - an ellipsis, one well-placed emoji - so replies do not read like an official document, but limit it to one emoji per message; more than two or three looks cluttered.
With a Growth-tier Zalo OA package (roughly 2,500,000 VND/year) or Comprehensive tier (roughly 6,000,000 VND/year) that includes API access for an AI agent, the automated reply scripts need to be loaded into the system as sample sentences in the agreed voice, with instructions for the AI to keep the same way of addressing customers even when a question falls outside the script. Full package details and how to connect an AI agent are in our Zalo OA chatbot for business guide.
Customers read your messages and feel the brand's voice right on their phone screen. Stock photo (CC0).
What voice mistakes make a Zalo OA chatbot lose customers?
The most common mistake is switching voice mid-conversation - the chatbot opens casually but suddenly turns formal in the order-confirmation message because that template was copied from somewhere else. The second mistake is overusing stock phrases like "Thank you for your trust and support" in every single message, which reads like a recorded announcement rather than a real person. The third mistake is going silent on off-script questions instead of replying "Let me check with the shop owner on that, one moment" - keeping the same voice even when admitting uncertainty. The fourth, less noticed mistake is letting staff answer manually in their personal voice instead of the brand's - each shift addressing customers differently, so a customer chatting in the morning meets one voice and a completely different one at night on the same Zalo OA page.
Zalo allows businesses to send advisory messages outside the 48-hour window for roughly 55 VND per message (Zalo Business Solutions, 2026 pricing) - meaning every message sent out has a real cost, which is one more reason to get the voice right so customers do not feel annoyed receiving it. For shops running both an automated chatbot and manual replies, the simplest fix is to pin the brand voice guideline right next to the screen where staff handle Zalo, instead of letting each person go from memory.
How do you check whether your chatbot's brand voice stays consistent?
The simplest method needs no special tools: once a month, randomly pick 10 recent Zalo OA conversations (both automated and manually answered), read them back, and ask three questions - is the way of addressing customers consistent throughout, did any banned word slip in, and if the shop's name were hidden, could you still guess which brand this is? If the answer is "yes" to all three, the brand voice is holding up well. Make this review a recurring operating habit - a one-time setup that gets forgotten lets the voice drift shift by shift. And when you upgrade from a scripted chatbot to a context-aware AI agent, the voice-keeping principles in our brand voice for AI agents guide carry over directly.
Frequently asked questions
How is brand voice different from brand tone?
Brand voice is a fixed personality that never changes - for example, always addressing customers the same casual way. Brand tone is the shade that shifts with context, upbeat when confirming an order, serious when apologizing. A chatbot must keep its voice fixed while adjusting its tone.
Does a Zalo OA chatbot need separate scripts for every situation?
Yes, but every script must follow one shared brand voice guideline (how to address customers, signature words, banned words). Write separate scripts for greetings, quotes, order confirmations, and out-of-stock apologies - but keep the same voice throughout, which is what many small shops overlook.
Which Zalo OA package is needed to run a chatbot with a defined brand voice?
You need a package with API access to connect an AI agent or automation system, usually starting from the Growth tier at roughly 2,500,000 VND/year (Zalo Business Solutions, 2026 pricing). The Standard tier at roughly 1,000,000 VND/year only supports manual one-on-one chat, no automated chatbot.
Do you need to hire a consultant to build a brand voice?
Not necessarily for a small shop. A 2-3 hour working session between the owner and whoever writes the scripts, agreeing on how to address customers, 5-7 signature words, and a banned-word list, is enough to start - then refine it gradually through real conversations with customers.