While everyone's been watching the chatbots, a quieter version of the same shift has been running in kitchens, cars, and pockets for years. Someone asks a speaker to find a pizza place that delivers late, or asks the car for a pharmacy on the route home, or asks their phone, hands full, whether the hardware store is open. The assistant answers out loud. Usually with exactly one suggestion.
Voice is the most unforgiving format in the visibility game, and the recent wiring of full AI models into these assistants has made it more capable and more consequential at the same time.
One answer, zero silver medals
A screen can hedge. It can show three options with a map and let the human choose. A voice answer mostly can't. Speaking five options at a driver is bad design, so assistants compress ruthlessly: one recommendation, maybe two, chosen with the same kind of record-reading the chat engines do.
That makes voice the purest version of the game. On a screen, being option three still gets you seen. Out loud, second place is silence. Whatever tips the assistant's confidence toward one business over another decides everything.
What voice queries actually are
Voice questions skew hard toward action and logistics. Is it open now. Which one is closest. Can I book for tonight. Call them. Directions. People talk to assistants when their hands or eyes are busy, which means they're disproportionately mid-errand and ready to transact.
Notice what that implies: voice answers lean heavily on operational data. Hours, location, phone number, booking availability, the word "open" being true when the assistant says it. The glamorous parts of marketing barely register here. The unglamorous parts are the whole contest.
Where assistants get their facts
Mostly from the same records the rest of AI search reads, with extra weight on the big structured sources: your Google Business Profile for anything in that ecosystem, Apple's business listings for Siri and CarPlay, Bing's data for the Microsoft world, plus the major review platforms layered on top for the "best" and "good" questions.
If you've been following the standard AI-visibility playbook, consistent listings, current hours, steady reviews, machine-readable website, you're already doing most of voice optimization. The difference is the penalty structure. A wrong closing time in a chat answer wastes a click. A wrong closing time spoken to someone already driving to you creates the angriest kind of customer, and assistants that burn users on bad data learn to route around the businesses that caused it.
The checklist, voice-flavored
Keep hours obsessively current, including holidays, on your Google and Apple listings especially. Make sure your phone number is right everywhere, because "call them" is a top voice action and it dials whatever the record says. Get bookable online if your industry supports it, since assistants increasingly complete reservations themselves. And say your area plainly in text, because "near me" is the most common voice modifier there is, and the assistant resolves it against your stated location and service area.
Then test it. Ask your phone's assistant and your speaker the questions your customers would ask, from your parking lot and from across town. What gets recommended, and is your data right in the answer? Five minutes, twice a year, and you'll know.
Voice never got the hype cycle the chatbots got, but it's been normalizing the core habit, ask a machine, act on the one answer, for a decade. The customers are already talking to the assistants. The assistants are already answering. The only open question is whether the record they read makes you the answer, and that part, fortunately, is yours to write.
About author
Sylas is the brains behind bold business roadmaps. He loves turning “half-baked” ideas into fully baked success stories (preferably with extra sprinkles). When he’s not sketching growth plans, you’ll find him trying out quirky coffee shops or quoting lines from 90s sitcoms.
Sylas Merrick
Head of Strategy
