Voice assistant
In one sentence
A voice assistant is a general-purpose spoken helper tied to a device or platform, such as those on phones and smart speakers, that works for the user across many tasks rather than for a single business.
Not to be confused with Voice agent.
Definition
A voice assistant is a general-purpose spoken helper tied to a device or platform, such as the ones on phones and smart speakers. It serves a user across many tasks rather than serving one business.
It is the thing that sets a timer, plays music and reads out the weather. It works for the person who owns the device.
The defining characteristic is allegiance. A voice assistant works for the user across an open-ended range of tasks. A voice agent works for a business within a defined scope. That single distinction explains most of the differences that follow.
What follows from serving the user rather than a business
- Breadth over depth. It has to handle weather, timers, music, messages and general knowledge, so it cannot be deeply expert in any one domain.
- Platform-owned. The device or operating-system provider builds and controls it, and a business reaches its users only through whatever integration the platform permits.
- Persistent identity. It is the same assistant across every interaction a user has, which supports personalization and raises the privacy stakes.
- Wake-word activated. Because it is always available, it needs a trigger phrase, and therefore some form of continuous local listening.
Why businesses have struggled to reach customers through them
- Platform intermediation. The assistant decides what to surface, so a business competes for that decision rather than owning the relationship.
- Discovery is poor. Users rarely install or invoke third-party capabilities, and several major platforms have pulled back their third-party developer programs.
- Attribution is invisible. A business cannot see what the assistant said about it.
- This is a direct parallel to the zero-click search dynamic, and it came first.
How the boundary is blurring
- General assistants now do things that once meant visiting a business, including bookings and purchases.
- Business voice agents increasingly handle open-ended questions rather than narrow scripts.
- The line holds on allegiance rather than on capability, and allegiance is the durable part.
Common misconception
The usual assumption is that deploying a voice agent means competing with the major voice assistants. They sit in different positions. A business voice agent is deep in one domain and owns the customer relationship. A general assistant is broad, shallow in any single domain, and mediates that relationship.
Why it matters commercially
The strategic lesson from a decade of voice assistants is that platform-mediated access to customers is fragile. A voice presence a business controls, on a property it owns, is a structurally different asset from a capability published into someone else's assistant.
Where AsqVox fits
The Orb belongs to the business and runs on the business's own site. That is the opposite position from publishing a capability into a platform assistant, and it is worth stating plainly because buyers conflate the two.
Visual
Who does it work for
The question is not what it can do. It is whose side it is on.
Statistics
Every figure carries its source and year. Vendor numbers are labelled as vendor numbers, and where no reliable figure exists this page says so rather than borrowing one.
There is no reliable, current, independent figure for voice assistant installed base or active usage. Platform-published numbers are marketing disclosures rather than audited figures, and several are years old.
-no reliable figureTreat any specific device or user count with caution and date it explicitly. A platform disclosure is not an independent measurement.
There is no reliable public data on third-party capability usage rates within the major assistant platforms, which is precisely the number that would substantiate the discovery problem.
-no reliable figureThe absence is itself telling. The figure that would prove or disprove the poor-discovery claim is the one the platforms do not publish.
Zero-click search reached 68.01 percent of US Google searches in early 2026, up from 60.45 percent in 2024.
68.01%independentSparkToro with Similarweb data, published June 2026, 2026 - US Google searches specifically, at a point in time. It is independent measurement of the same intermediation dynamic voice assistants exhibited first, and it is far better evidenced than any assistant-usage figure.
Independent latency testing put realtime model time to first token from around 0.78 seconds for xAI Grok Voice to around 2.98 seconds for Gemini 3.1 Flash Live.
0.78s to 2.98sindependentArtificial Analysis via Softcery, April 2026, 2026 - A capability reference point for the underlying models. Time to first token is a component of the spoken response budget, not the full time to first audio a listener hears.
Examples
In practice
A retailer invests in publishing a capability into a major voice assistant platform. Usage stays minimal, because users do not discover or invoke it. The retailer cannot see what the assistant tells people about its products, and a platform policy change alters the integration twice in two years. The same budget spent on on-site voice produces interactions it can measure, attribute and control.
The everyday version
A voice assistant is the thing on your phone or speaker that sets a timer and tells you the weather. It works for the person who owns the device. A voice agent on your website works for you. They both talk, but only one of them is yours.
Usage
Who says it
- Consumers and general media use it to mean the major platform assistants.
- Product strategists use it when they are discussing platform dynamics.
- Vendors use it loosely to describe business voice agents, which muddies the distinction.
Where it turns up
- In consumer technology coverage, platform strategy documents and device specifications.
- Rarely in business voice procurement, except as a comparison point.
Common misuse
- Describing a business voice agent as a voice assistant. It invites the wrong mental model: breadth and platform ownership.
- Assuming assistant-platform reach is available to a business. Discovery and attribution have both proven weak.
- Citing dated platform usage figures as if they were current.
Questions people ask
What is a voice assistant?
A voice assistant is a general-purpose spoken helper tied to a device or platform, such as a phone or a smart speaker. It works for the user across an open-ended range of tasks, from timers to weather to general knowledge, rather than for a single business.
What is the difference between a voice assistant and a voice agent?
Allegiance. A voice assistant works for the user across open-ended tasks and is owned by the device or platform. A voice agent works for one business within a defined scope and is owned by that business. Capability between the two is converging, but allegiance is the durable distinction.
Does deploying a voice agent mean competing with the major voice assistants?
No. They occupy different positions. A business voice agent is deep in one domain and owns the customer relationship. A general assistant is broad, shallow in any single domain, and mediates that relationship. A voice presence a business owns on a property it controls is a different asset from a capability published into someone else's assistant.
Why have businesses struggled to reach customers through voice assistants?
Because the platform intermediates. The assistant decides what to surface, third-party capabilities are rarely discovered or invoked, and a business cannot see what the assistant said about it. That is the same intermediation dynamic zero-click search shows, and voice assistants exhibited it first.
Last reviewed 31 July 2026. Written and reviewed by Dhruv Dholakia, founder of AsqVox.