How to teach a parent or grandparent to use ChatGPT safely

By Gaone Dixon, 29 October 2026, 9 min read

The short answer: set ChatGPT up on your parent's own phone or tablet, under their own account, from the official app. Secure the sign-in together, switch off model training if they'd like more privacy, and agree three rules before the first real question: never type anything private, check anything that matters, and never act on an urgent call or message without hanging up and checking. Then teach in three short sessions where they hold the device and you only coach. Most of the safety comes from the setup and the rules, not from how much they learn on day one.

This guide is written for the person doing the teaching. If your parent or grandparent would rather learn on their own, send them our calm first week for beginners over 50 instead; this article doesn't repeat its seven-day plan. Every claim about how ChatGPT works links to OpenAI's own help pages, and the scam advice links to the FBI, all fetched on 29 October 2026.

Should they use their own account, or mine?

Their own. It's tempting to sign them into your account to save time, but it causes three problems:

Mixed history. Chat history syncs across every device signed in to the same account, according to OpenAI's iOS and Android app FAQ. Your parent would see your chats and you'd see theirs.
Mixed memory. When memory is on, OpenAI's Memory in ChatGPT page says ChatGPT can use past chats and saved details to personalise answers. Your work, your children and their health questions would all end up in one profile.
Their safety habits never form. People learn the settings by changing them on their own account. If you own the account, they never do.

A separate free account fixes all three. If they'd like you to be able to help later, write the account email in their notebook, not the password.

How do I set it up safely on their device?

Do this together, with them pressing the buttons. Allow about 20 minutes.

Install the real app. OpenAI's app FAQ says to check that the developer is OpenAI OpCo, LLC on the App Store and the publisher is OpenAI on Google Play, and to search for "OpenAI ChatGPT" rather than "ChatGPT". Check the name under the app before tapping Install. On a computer, type chatgpt.com directly rather than clicking a search ad.
Use an email they can get into. If they can't open their own inbox, fix that first: they'll need it to sign in.
Add a second sign-in step. OpenAI's guide to managing multi-factor authentication lists the options: an authenticator app, push notifications, a 6-digit code by SMS or WhatsApp, or a passkey. For most older beginners, a text-message code is often the easiest to understand. Turn it on in Settings > Security and login.
Set up a backup way in. The same page warns that email isn't an MFA method, and that if every method is lost, OpenAI Support offers a one-time recovery that "can take a few days" and "is available only once". Add a second method (for example, a passkey on the phone as well as SMS) and note it in their notebook.
Decide on model training together. OpenAI's data controls page says that on free and personal plans "data sharing is enabled by default". To switch it off: Settings > Data controls > Improve the model for everyone. Chats stay in their history either way. One catch worth explaining: giving a thumbs up or down can let the whole conversation be used for training, even with the setting off.
Show them where memory lives. Settings > Personalization > Memory. They can also simply ask ChatGPT what it remembers about them. Leave it on or off, but make sure they know where the switch is.
Write it down. Which app, which email, which sign-in methods, which settings you changed. Our what not to tell ChatGPT guide covers the settings in more detail if they want to go further.

What three rules should I agree with them first?

Keep it to three. People remember three rules; they forget twelve.

"Nothing private goes in." No passwords, bank codes, card or ID numbers, or other people's details. If they need help with a letter, they describe it in general terms. The full list, with placeholders to use instead, is in what not to tell ChatGPT; print it and put it by the device.
"Check anything that matters." OpenAI's own Voice help page says "ChatGPT can make mistakes" and to check important information. Health, money, legal and travel answers get checked with a pharmacist, the bank's own site or an official source. Our guide on how to check an AI answer has a five-minute routine.
"Urgent means hang up and check." This rule is about scams, not ChatGPT itself, but it belongs here because AI makes scams more convincing (see below).

Ask them to say the rules back in their own words. If they can explain them to you, they'll use them.

How should I actually teach it?

The most useful thing you can do is keep your hands off the screen. If you take the phone and "just do it", they'll need you again next time. Some habits that help:

They hold the device, you sit beside them. Point and describe ("the round button at the bottom right") instead of tapping.
Start with something they already care about, not a demo of what AI can do. A recipe, a garden question, a letter they've been putting off.
Keep each session short: 20 to 30 minutes, then stop while it's still going well.
Let them make the mistakes. A wrong tap is fine; nothing breaks. Showing them how to start a new chat is more useful than preventing every slip.
Have them write their own notes in their own words. Your notes are less use to them than theirs.

A simple three-session plan:

Session 1 (setup): install, sign in, second sign-in step, settings, the three rules. One easy question at the end so the session finishes with a win.
Session 2 (talking): two or three real questions and follow-ups ("make that shorter", "explain the second part again"). Try Voice if typing is a struggle.
Session 3 (safety practice): rewrite a private question with placeholders, check one answer against an official source, and rehearse what they'd do if a "grandchild" called asking for money.

After that, step back. Offer a short check-in a week later rather than daily help.

Is talking to it easier than typing?

For many people, yes. OpenAI's ChatGPT Voice page explains that Voice lets you "speak with ChatGPT and hear a spoken response", with the text shown in the chat so you can read along. A few details from that page are worth knowing before you teach it:

Starting it: tap the Voice icon in the message bar and allow the microphone when asked. They can choose a voice they find easy to hear.
Interruptions: background noise or long pauses can make it cut in. Headphones or a quieter room help, and they can say "wait until I ask you to respond" at the start.
Transcripts aren't exact. OpenAI says Voice transcripts "may not exactly match" what was said, so anything important should be read back.
Audio is stored with the chat. The page says audio clips from Live and Advanced Voice are kept with the transcript for 30 days, and are not used for training unless the user chooses to share them.

If spelling or small keyboards are the barrier, Voice can make the difference between giving up and carrying on.

How do I explain the scam risk without scaring them?

Keep it factual and short. The FBI's public service announcement Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (December 2024) warns that criminals use AI to write more believable messages, create fake profile photos and generate "short audio clips containing a loved one's voice" to fake a crisis and ask for money.

The FBI's own tips map neatly onto rule 3:

Agree a secret word or phrase with the family, so a caller can prove who they are.
Hang up and call back on a number you already have or have looked up yourself.
Never send money, gift cards or cryptocurrency to someone you've only met online or by phone.

Two ChatGPT-specific points, both from OpenAI's Verifying Communications from OpenAI page: OpenAI says it will "never go into details of your account over social media", and lists its real email domains (such as @openai.com and @email.openai.com) and apps. Anyone messaging your parent offering to "fix" their ChatGPT account is a scammer.

The family code word is set up step by step on Day 7 of our over-50 first week, so you can do it together at the end of Session 3.

Copyable prompts for the first sessions

Have your parent type or say these themselves. Swap the bracketed parts.

First conversation (Session 1):

I'm new to ChatGPT and I'm [age range, e.g. in my seventies]. Please answer in short, plain sentences, one step at a time, and ask me if I'd like more detail rather than giving me everything at once. My first question: [their question].

Practising placeholders (Session 3):

I've had a letter from [my bank] saying [general description, e.g. my account will be "restricted" unless I call a number]. I won't share any personal details. Is this the kind of message banks normally send, what are the warning signs of a scam, and how should I check it safely?

Checking an answer (Session 3):

You just told me [the claim]. How confident are you? What would I need to check, and which official source should I check it with? Please say plainly if you're not sure.

Frequently asked questions

Is ChatGPT free for older people?

There's a free plan, and the setup in this guide works on it. OpenAI's Voice page shows the free plan has limited Voice use, with limits that "may change". Nobody needs to pay to learn the basics.

Can I see what my parent asks ChatGPT?

Not with separate accounts, and that's the point: it's their private tool. If they want help, they can show you a chat on their screen. Sharing one account so you can watch would mix your histories and memories.

What if they forget their password or lose their phone?

That's why step 4 above matters. With a backup sign-in method written down, they can get back in. OpenAI's MFA page says that if every method is lost, recovery through Support takes a few days and is available only once.

Should I turn memory off for them?

It's a choice, not a safety requirement. Memory makes answers more personal, but it also means ChatGPT keeps details. If they'd prefer a clean slate each time, switch it off in Settings > Personalization > Memory; OpenAI says turning it off doesn't delete past chats.

What if they've already typed something private?

Delete the chat (open it, tap the ••• menu, then Delete). If it was a password or bank detail, change it with the bank or service straight away; deleting the chat doesn't undo sharing. Our what not to tell ChatGPT guide covers this step by step.

Not ready to buy? Start with the free guide, Your first week with AI: ten prompts and a privacy checklist, plus a five-day email course.