Talk to AI: get more from voice mode
Talking to AI is easier than typing when your hands are busy or your thoughts are messy. Here’s what voice mode is good for, how to steer it and where it slips.
In 30 seconds
- Use Jargon busterVoice mode: A way of using an AI assistant by talking to it and hearing it answer out loud, instead of typing and reading. to rehearse, think out loud or cook hands-free. Never handle your phone while driving.
- Steer it like a conversation: interrupt, ask for shorter answers, and say “one question at a time”.
- It can mishear names and numbers, and still be wrong. Check your privacy settings, too.
With flour on your hands, out on a walk, or rehearsing what you’ll say to your landlord, typing gets in the way. Voice mode lets you talk to an AI assistant and hear it answer out loud. A few habits make it far more useful, and a few limits are worth knowing.
Many assistants have one, including ChatGPT, Gemini, Claude and Microsoft Copilot. Don’t confuse it with dictation: in many apps the microphone button only types what you say. Voice mode is usually a separate button, and it talks back.
What it’s good for
- Rehearse a conversationAsking for a pay rise, returning something faulty, a job interview: give it a role and practise out loud.
- Think out loudTalk through a messy problem on a walk, let it ask you questions, then ask for a written summary at the end.
- Cook hands-freeAsk for one step at a time, say “next” when you’re ready, and convert grams or oven temperatures as you go.
- Show it what you seeSome apps let you switch on the camera mid-chat. Point it at a plant, a menu or your washing machine’s dials, and ask.
That last one works because most assistants are Jargon busterMultimodal: An AI that understands more than one kind of input, such as text, pictures, sound and video.: they can take in pictures as well as words. And if you’re learning a language, voice mode makes a patient speaking partner: see learn a language with AI.
Steer it like a conversation
Voice mode works best when you treat it like a person on the phone. Give it a short brief at the start, and redirect it as you go. In many apps you can simply start talking to interrupt; in others, you tap the screen.
You: Play my manager. I’m going to ask for a pay rise. Keep your replies short, and ask one question at a time.
AI: Okay, I’m your manager, and I’ve got ten minutes before my next meeting. What did you want to talk about?
You: Let’s stop there. How did my opening sound?
AI: A little apologetic. Try leading with what you’ve achieved this year, then make the request plainly.
| If it… | Try… |
|---|---|
| Talks for too long | Cutting in: “Stop. Two sentences, please.” |
| Fires several questions at once | “One question at a time.” |
| Reads out a long list | “Just the first step. Wait until I say next.” |
| Gets a name or number wrong | “Spell that back to me.” Or: “Read the number back, digit by digit.” |
| Jumps in while you’re thinking | Muting the microphone while you think, if your app has a mute button |
For this chat I’m listening, not reading. Keep each answer to two or three sentences. Ask me one question at a time and wait for my answer. Don’t read out links, tables or long lists. If you’re not sure you heard a name or a number correctly, check it with me before you carry on.
Where it slips
Jargon busterSpeech recognition: Technology that works out which words you’ve said, so a computer can respond to speech. It’s behind dictation, voice assistants and AI voice modes. still stumbles over names, numbers, accents and background noise, so “fifteen” can become “fifty”, and Siobhan can become someone else entirely. Ask it to read back anything that matters. Many apps also save a Jargon busterTranscript: A written, word-for-word record of what was said in a recording, a call or a meeting. of the chat, which you can check afterwards.
A calm, confident voice can make a wrong answer sound right, and there’s often nothing on screen to check. It can still state something false as fact, a Jargon busterHallucination: When AI states something false as if it were true, because it predicts plausible words rather than checking facts.. For anything important, switch to text and check the answer properly. For health, money or legal decisions, speak to a professional.
Check what it keeps
Depending on the app, your voice chats may be saved as text and sometimes as audio, and may be used to improve the company’s models unless you switch that off. Look in the privacy or data settings, and delete what you don’t want kept. Remember that people nearby may be picked up too, and don’t say anything you wouldn’t type, such as passwords or account numbers.
Check yourself
3 quick questions nothing is savedTools in this guide
- CChatGPTOpenAI’s general-purpose assistant for writing, questions, analysis and images.
- GGeminiGoogle’s assistant, built into Gmail, Docs and Android.
- CClaudeAnthropic’s assistant, strong at long documents, careful writing and code.
- MMicrosoft CopilotMicrosoft’s assistant across Windows, Edge and Microsoft 365 apps.
Spotted a mistake? Tell us and an editor will check it.
