Voice on the go: think out loud on the pram walk

The situation

One hand on the pram, the other holding a scooter, and your best thinking happens exactly now, when you can’t type. By the time you’re home, half of it is gone. Your voice can catch it instead.

By the end of this lesson you’ll know the two ways to use your voice with an AI app, which one to pick for what, where the buttons are, and how to come home from a walk with a sorted list or a plan for a tough evening.

Why it’s worth it

A lot of the mental load gets processed in the gaps: walking to school, waiting at the swimming pool, pushing a pram round the park. Those are the moments your brain finally has room to think, and they’re exactly the moments your hands are full. Typing a long list on a phone with one thumb is slow and fiddly, so most of those thoughts never get captured.

Speaking is much faster than thumb-typing, and it works with your hands busy. The AI apps from Lesson 1.1 can take what you say, turn it into text, and then do the same sorting and drafting they do with typed messages. The trick is knowing which kind of voice feature to use, because there are two, and they’re good at different things.

Two ways to use your voice

Dictation turns your speech into text, and that’s all it does. You tap the microphone in the message box (or on your phone’s keyboard), talk, and your words appear as a written message. You can glance at it, fix anything, and send it like any other message. The AI then answers in writing. Dictation is best for brain-dumps and lists, because you get everything out in one go and the result is written down in the chat for later.

A voice conversation is a spoken back-and-forth, like a phone call. You talk, the AI answers out loud, you reply, and so on. It’s best for thinking something through, when you want a response as you go: venting, deciding between two options, rehearsing a tricky conversation. Because it answers as you go, it’s less good for a long list: it tends to respond to each bit before you’ve finished.

Each app has its own button for a voice conversation (names and icons change between versions):

• ChatGPT: the Voice button in the message bar.
• Claude: the sound-wave icon next to the microphone in the message box.
• Gemini: Live, at the bottom of the app.

The first time you use either feature, your phone asks for permission to use the microphone: allow it. Free plans include voice, with usage limits that are often tighter than for typing. If you hit a limit, switch to dictation, which sends ordinary text messages.

	Dictation	Voice conversation
What happens	Your speech becomes a written message you send	You and the AI talk back and forth out loud
Best for	Brain-dumps, lists, long rambles	Venting, thinking aloud, rehearsing what to say
What you end up with	Your words and its answer, in writing	A spoken chat (ask for a written summary at the end)

Why it mishears, and where

Before the AI can understand you, your voice has to be turned into words. That step works by matching sounds to the most likely words, which is why it’s usually impressively accurate and occasionally hilariously wrong. It slips in predictable places: numbers that sound alike (“fifteen” and “fifty”, “thirteen” and “thirty”), names it hasn’t heard often, and anything said over wind, traffic or a crying toddler.

The AI then reads the words it was given, not what you meant. So if “fifty” became “fifteen”, the sorted list will say fifteen, with total confidence. Strictly speaking that isn’t guessing: the AI is faithfully working from a wrong starting point. But the result looks exactly like the confident guessing from Lesson 1.1, a wrong detail stated as fact, and the fix is the same: treat numbers, names and dates that came in by voice as “probably right”, and check them before they matter.

One more thing about voice and privacy: voice chats are stored with your other chats, so the never-share list from Lesson 1.4 applies to what you say just as much as to what you type. And the people around you can hear you too, so the school gate isn’t the place to dictate your child’s medical appointment.

Copy this

For a brain-dump (dictation). Tap the microphone and say this, with your own ramble in the middle:

Brain-dump coming, it's messy. [Talk for as long as you need: everything on your mind.] Okay, that's it. Sort it into today, this week, can wait, and someone else's job. Don't add anything I didn't say. Ask me up to three questions at the end.

What you’ll see: your words appear in the message box as you speak, with the odd mistake. Send it and you get the same sorted list as in Lesson 1.1.

For a hard evening ahead (voice conversation). Start a voice chat and say:

I've got a tough evening coming: [what's ahead, e.g. homework battles, a late work call, two overtired kids]. Let me vent for a minute without you fixing anything. Then help me pick one thing I can drop tonight, one thing to prepare now, and one calm sentence I can say when it all kicks off. At the end, write the plan in the chat as a short list.

What you’ll see: it listens, answers out loud, and once you’re done, the plan appears in the chat as text.

Why each part is there

“Okay, that’s it” in the dictation version marks the end of your ramble, so the instructions after it don’t get mixed in with your to-dos.

“Let me vent without you fixing anything” matters because AI apps are eager to help, and their idea of helping is solutions. Sometimes you need to get it out first. Saying so up front means it listens and acknowledges rather than jumping straight to tips.

“One thing to drop, one to prepare, one sentence” keeps the plan small on purpose. When you’re tired, a detailed schedule is one more thing to carry. Three small, concrete decisions are doable, and they’re usually enough to change how the evening feels.

“Write the plan in the chat” solves the one weakness of voice conversations: spoken answers disappear. A short written list is something you can glance at when you walk in the door.

What you’ll get

An example of the plan (yours will be worded differently):

That’s a lot to carry into one evening. Here’s a plan:

• Drop tonight: the reading log. One night off won’t matter; a calmer bedtime will.
• Prepare now: fill the pasta pan before your call, so dinner is ten minutes away when you finish.
• Your calm sentence: “You’re really tired. We’re just doing the next small thing: teeth.”

Notice that it acknowledged the load before planning, which is what the “let me vent” line asked for. Each item is small and specific: not “simplify dinner” but “fill the pasta pan before your call”. And the calm sentence is short enough to remember mid-meltdown. If any of it doesn’t fit your family, say so (“We don’t do reading logs, try something else”) and it adjusts.

Before you rely on it

Voice is brilliant for getting things out of your head, less so for details. Before any of it goes into a calendar or a message, check:

• Numbers. Amounts, ages, quantities. “Fifteen” and “fifty” sound alike, so check against the original letter or message.
• Names. Unusual names often come out as ordinary words (“Saoirse” can become “sir sha”). Fix them before anything gets sent to anyone.
• Times and dates. “The twelfth” can become “the twentieth”, and “half five” (5.30) can confuse it. Check before it goes in a calendar.

If you spot a slip, just say it: “Not fifteen, fifty.” It corrects the list, because corrections are just more context.

Make it yours

• Remind me what I said I'd do today, as a list. Handy at the end of a long voice chat, when the actions are buried in the conversation.
• Say that again in two sentences. Spoken answers can run long; this trims them.
• Answer in [your family's other language]. Useful when you think in one language and the school writes in another.
• Change the voice it speaks with in the app’s settings, if the default one grates on you. It’s a small thing, but you’re more likely to use a voice you like.

If it goes sideways

• It keeps cutting you off. In a voice conversation, the app decides you’ve finished when you pause, and pauses are normal when you’re thinking. Say “I’m not finished, wait until I say done”, or use dictation for long dumps.
• It mishears everything. Wind and traffic are hard work for any microphone. Talk in shorter bursts, hold the phone a little closer, or turn your back to the wind.
• You hit a voice limit. Voice features often have tighter free-plan limits. Switch to dictation: it’s just text, so it carries on working.
• It gave you a full plan anyway. AI apps sometimes over-deliver. “Just the three things, please” brings it back.