How to write for voice assistants.
When writing for voice assistants, you must keep in mind that you are delivering a simultaneous mix of content, interactive rules, story-flow feedback, brand experience, and voice-driven design. A mix designed to support human engagement through a conversational UI model. But to help you better understand writing for voice assistants, first things first…
What’s a conversational UI for voice assistants?
A conversational user interface (Conversational UI) is a digital experience around your product or service that creates the feeling of having a back-and-forth conversation between your device and a real human being.
It’s a simulated interaction. Your device’s or service’s responses are triggered by a machine’s understanding of parsed phrases it hears and analyzes from the user. That, coupled with the machine’s ability to extract keywords and set rules and relationships around those keywords, a voice device can calculate the probability of a specific user’s intent or request.
All that translated into English: it understands what the user is asking for.
Once the intent is established, how your device delivers a vocal response (along with its ability to satisfy the request), is what helps humans feel as if they are having an organic and productive conversation. That’s the “conversational” part of “conversational UI.”
Now let’s talk about the “UI” part for voice assistants.
The UI (user interface) refers to the medium(s) employed to manage that back-and-forth engagement. For a web site or most standard mobile experiences, it’s a graphic interface. That includes things like the images you see and the buttons you click on.
With voice assistants, a conversational UI can come a couple of ways.
One way is as a hybrid conversational interface (CI). This is where voice is used in coordination with a graphic user interface (think how Siri or Google Assistant simultaneously present voice and visual displays in their response on your smartphone. The smartphone voice may say, “Here’s today’s weather,” while the screen pulls up an image of the weather forecast.
A conversational UI can also be voice only.
Think Alexa in the Amazon Echo. In this voice assistant product, there is no graphic display. All interaction with the user is through voice.
In voice-only devices and services, you are writing for an audio-based UI. The voice and messaging you write also serves as the medium by which humans understand how to engage with your platform.
That means the voice that you write for in a voice assistant is doing what the graphic interface would be doing. So..
No button taps. Instead, you write to create spoken words that, like buttons, request action. “Play” “Open” “Stop.
No visual-based status. Instead, you create words that give vocal confirmations. “Playing song.” “Unable to find that information.”
You are delivering content. That can be a song. An answer about the weather. Or in a hybrid CI, you are framing the graphic content being presented so the user can better understand what information to consume.
Your writing for voice assistants is also about brand.
Like any other UI, as a writer, you are also a participant in the delivery of an experience. As any experience creates or reinforces a brand, the words you craft around voice are also part of the brand.
If radio is called theatre of the mind. Voice UI is the digital interface of the mind.
In my work on voice for Comcast and well as a product currently in development, I’ve learned some things to keep in mind when writing and creating content for voice assistants – as well as the people those assistants serve: humans.
1. Be brief and relevant.
You’re not trying to be a buddy. Don’t be overly chatty. Stick to proving your worth by being an assistant that provides value. That means being laser-like and focused on delivering content users want as efficiently and as fast as possible.
2. Write in clear, simple language for voice.
Try to write so your responses are purposeful, confident, responsive, helpful and insightful.
Some examples:
With Purpose: “How can I help you?”
Confidently: “I’ll get those forms for you.”
Helpful: “Your subscription is expiring on May 8. Should I renew it for you?”
Insightful: “I see you’ve got a meeting at 3:00 PM with Bob Smith. Would you like to see Bob Smith’s profile?”
3. Don’t make your assistant a liar.
Don’t be lazy or make overstatements with writing for voice response. Doing so can produce logic holes or statements that can be proven inaccurate or untrue. For example:
VOICE: “I’m sorry, you don’t have any records on X.”
Great! If that’s absolutely true.
But companies make mistakes. A relevant record might be misplaced, mis-tagged or not entered yet. This can lead to existing data that your user wants but your voice assistant doesn’t know about. Yet its response has it speaking in absolutes.
Remember, your written voice is an audio UI. So imagine how a user would feel if a visual UI, gave a bad number. Would you want a speedometer in the car you were driving to be “kinda-sorta” accurate?
Discovering that your voice assistant gives information that’s up for suspicion can ruin user trust for accepting any future information.
Make sure you write responses to ensure what your voice assistant says is indisputably true – or cannot be directly refuted. If there is a possibility of an exception to your voice statement, conditionalize your voice response.
So instead of:
VOICE: “You don’t have any records on X.”
It becomes…
VOICE: “I can’t find any records related to X” or “I’m currently unable to find any records related to X.”
Think of your voice response as a spokesperson for your service or company. As the voice representing the organization, you have to think about the company’s needs while still being an honest broker of information in order to keep user trust in your product.
4. Make sure you don’t overshare.
Reading doesn’t make noise. Unlike text, people can hear information given via a voice response. That includes sensitive data. Stories, where this will be an issue, could include regulated markets like financial or health data. If you are giving such information, via voice assistant, do not use voice to deliver it unless specifically prompted to do (and within your organization’s rules for sharing).
4. Your voice is also brand.
Most of us know Siri’s voice when we hear it. Same with Alexa. Same with the famous but fictional HAL 9000 from 2001: A Space Odyssey. Most of us would think of Siri as cheerful and confident. While anyone who has heard the HAL 9000s voice has experienced a slight shiver from its calm, detached creepiness.
Voice is brand. For companies and products using voice to deliver services and content, it will become the audio brand. So consciously write with the style, tone, and demeanor you wish to reflect that brand.
“Yo, what up?” Is very different than, “Good morning. How can I help you?” And those are very different than the memorable, “What do you think you’re doing, Dave?” from the HAL 9000.
5. “Sorry” is not a synonym for “no.”
When you have to tell a user that the sum of a request if zero (e.g. “There are no books by that title.”), don’t reflexively use “sorry” to start the sentence.
Sorry is not about your response being X=0. Sorry should generally be reserved for instances where you want your voice assistant (and by extension, the brand) empathizing with your user’s situation. Particularly for circumstances where you know the user may feel disappointment in the experience they expected. That includes a system error or a platform error.
5. Plan for Easter eggs/small talk.
When the user takes a moment of curiosity to play your device or service, it’s a good opportunity to have brand-reinforcing moments. That includes chat that can deliver lighter moments for company connection. Write a few jokes to be ready if the user asks for one. Or put some witty rejoinders in your voice assistant’s quiver to be ready for a user’s smart-alec questions.
Your voice writing: Be helpful. Be human. Be accurate.
The voice you write will define the experience users have with your digital experience. Align with these voice principles and you’ll get there.