voice assistant

in English: voice assistant

basicwidespread

A program in a phone, speaker, or car that you speak to, which responds verbally or performs what you requested.

In plain terms

A voice assistant is like a helpful housekeeper always within earshot: you do not need to approach it; simply address it. You state what you want, and it responds or handles the task for you.

It is especially useful when your hands or eyes are occupied: while driving, cooking, or bathing a child. It sets timers, starts music, allows you to dictate messages, tells you the weather, and turns off smart lights.

Older voice assistants only recognized pre-defined request types and easily got stuck if you phrased things differently. Newer ones are powered by large language models, so you can converse more freely, ask follow-up questions, and they understand more complex requests.

What primarily distinguishes a voice bot from a chatbot is the channel: here you do not type and read, but speak and listen. This makes it more convenient but also less accurate, because mishearing is a separate source of error. The line is increasingly blurred, as many chatbots can now be used for live spoken conversation.

How it works

Listening: the device's microphone continuously waits for a single wake word (such as 'Hey Siri' or 'Alexa'), which is typically recognized on the device itself.
Speech recognition: the sentence spoken after the wake word is converted into text by the system – either on the device or on the provider's servers, depending on the manufacturer and device.
Interpretation: the system – in newer assistants, a large language model – works out what you actually want: whether you expect an answer to a question or something needs to be done.
Execution: the assistant finds the answer or sends an instruction to an application or device (calendar, music player, smart home).
Answer: the finished text is read aloud by a machine voice, meaning speech synthesis converts it back to sound.
In the latest systems, these steps partially merge: the model processes audio directly, enabling faster and more natural responses.

An everyday example

While driving, you remember you need to call the mechanic. You do not reach for your phone; you simply speak to the car's voice assistant, say who to call, and it dials while your hands remain on the wheel.

Who is affected

everyoneIt is built into most smartphones, smart speakers, TVs, and new cars, even if you never consciously turned it on.

parent, student, teacherChildren pick up how to use it easily, so you should know what the home device hears, what it can buy and what it responds to. → Education & Family

entrepreneur, executiveVoice-based AI increasingly answers calls at telephone customer service centers, which reduces costs but can lose customers if poorly configured.

patient, doctorFor the visually impaired, people with limited mobility, and the elderly, voice control is often the only comfortable way to access digital devices. → Health

A common misconception

MythThe voice assistant continuously records every word you say and sends it to the manufacturer.

In realityBy default, it only listens for the wake word and forwards the request following it. False activation can occur, in which case a conversation segment not intended for it may be recorded.

Limits and risks

May mishear: in noise, with dialects, rare names, or multiple speakers, it often understands something different from what you said.
Support for the Hungarian language varies by provider and is often incomplete, so check before purchase whether the specific device supports Hungarian.
Some of your recordings end up on the provider's servers, so it is worth reviewing privacy settings and stored history.
A language model-based assistant can say false things in a confident tone, and in spoken form this is harder to notice than in writing.

Similar, but not the same

chatbot – You typically converse with a chatbot in writing and with a voice assistant orally. The same language model may power both, and many chatbots now offer a voice mode.
speech recognition – Speech recognition only converts sound into text; a voice assistant also interprets this and acts or responds.
AI-agent – An agent independently carries out multi-step tasks, while a traditional voice assistant fulfills individual requests; the two are increasingly merging.

Related concepts

Broader

virtual assistant, artificial intelligence, Human–machine interaction

Narrower

smart speaker, voice-based customer service, wake word

On the edge

chatbot, speech recognition, speech synthesis, large language model, AI-agent

A brief history

The masses encountered it in 2011, when Apple introduced Siri in the iPhone 4S. In 2014, Amazon Alexa brought it into living rooms with the Echo smart speaker, and in 2016 Google Assistant appeared. From 2025, major manufacturers began replacing old systems with successors built on large language models: Amazon introduced Alexa+, and Google is replacing Assistant with Gemini.

Where things stand

By the end of 2026, it is present in almost every smartphone and many household devices, with manufacturers replacing old command-based systems with conversational versions built on large language models. In Hungarian, the offering is still narrower and more uneven than in English.

Status: 2026 November

In the news

Appears in 2 Agent Age stories; the curve shows weekly occurrence in the news magazine.

K&H's Kate assistant interprets customer requests in conversation and can initiate transfersOct 21, 2026 · Money
K&H prepares to accept payments initiated with Mastercard Agent PaySep 22, 2026 · Money

Also known as

hangasszisztens · voice assistant · voice-controlled assistant · virtual assistant · digital assistant · voice assistant · virtual assistant · AI voice assistant

Sources

Wikipedia: Siri

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.