Is it safe to give your mom ChatGPT? Yes — with two rules and one card on the desk
Giving an older parent ChatGPT is safe when two rules are followed: never paste in personal financial or medical information, and use the free version unless they specifically need the paid one. The bigger risk is AI-imitation scam calls, which are a separate problem tracked by the AARP.
The reason you are asking this question — and the reason I have been asked it, in some form, by maybe a dozen people in the last year — is not really about ChatGPT. It is about the fact that your mother is in her seventies, and the world keeps inventing new ways for people to be tricked, and you have spent enough hours on the phone with her about the last thing to want to be sure about this thing before you put it in front of her.
I will give you the short answer first, and then the long one. The short answer is yes. ChatGPT is one of the safer tools you can put on her desk this year, in the specific sense that it does not call her, does not text her, does not autoplay anything, does not ask for her Social Security number, and does not have a comments section. What it does have is two failure modes a thoughtful gift-giver can close off in about ten minutes. That is what the rest of this piece is for.
The honest answer and the one risk worth worrying about
The thing people are usually picturing when they ask whether ChatGPT is safe for an older parent is something like a phishing attack or an account hack. That is not, in practice, where the risk lives. ChatGPT itself does not initiate contact with anyone. The account sits there. You open a tab, you type a question, you read the answer, you close the tab. Nobody emails her later asking for money. Nobody calls. The product, narrowly considered, is closer in surface area to a calculator than to a social network.
The genuine risk — the one I would actually worry about if my own mother were the one signing up — is something the AARP has been tracking for about three years now and which has nothing to do with ChatGPT specifically. It is the rise of AI-generated voice scams. The AARP's 2025 fraud-watch report flagged AI voice cloning as one of the fastest-growing scam categories targeting people 65 and older: a stranger calls, plays a thirty-second clip that sounds like your grandson's voice saying he's in trouble, and asks for a wire transfer. The Federal Trade Commission published a consumer advisory on the same pattern and noted that the technology required to clone a voice from a thirty-second sample is now widely available.
The connection to your question is this: AI voice scams are not caused by ChatGPT, and giving her a ChatGPT account does not make her more vulnerable to them. But because the words "AI" and "scam" are now adjacent in the news, many adult children — and many older adults themselves — have conflated the two. They are different problems. ChatGPT is a tool she controls. A scam call is a thing that happens to her. The right response to the second one is a family code word and a rule about never sending money based on a phone call, which is a separate conversation worth having on its own. The right response to ChatGPT, for the reasons I am about to lay out, is much smaller.
The two rules to write on the inside cover
Most of the safety work is handled by exactly two rules. If your mother follows these, the day-to-day risk profile of her using ChatGPT is roughly equivalent to the risk profile of her using a word processor.
Rule one: nothing she would not write on a postcard. That is the line. She can ask ChatGPT to help her draft a thank-you note, simplify an insurance letter, prepare questions for her cardiologist, swap an ingredient in a recipe, write a eulogy, decode a confusing email from the utility company, or rephrase a difficult message to her sister-in-law. What she should not paste into the box is her Social Security number, her bank account number, her Medicare ID, the password to anything, her full medical record, her home address, or anything she would not be comfortable having a stranger read aloud at a kitchen table. The reason is not that ChatGPT is going to steal it — OpenAI's stated privacy policy is reasonable and the company is a serious U.S. business — but that any information typed into a cloud service is, in principle, information that could be exposed by a future breach or a future policy change. The postcard rule is the cleanest mental model for an eighty-year-old who has never had to think about this distinction.
Rule two: when in doubt, don't act on it without checking. ChatGPT is occasionally wrong. Not catastrophically, not maliciously, but in the specific way that a confident-sounding paragraph can contain a wrong date, a wrong dosage, a wrong fact. For most uses — writing a letter, drafting a question, simplifying a paragraph — being wrong does not matter much, because she will see the output and either like it or rewrite it. For a few uses, it matters a great deal. The rule on the card should read: if the answer involves a medical decision, a legal decision, or money moving, call the doctor or the lawyer or the bank before acting on it. ChatGPT is a drafting tool, not an authority. The same way she would not refill a prescription based on something the lady at the grocery store said.
Those are the two rules. Everything else is texture.
The card itself matters. I have watched my father, twice, encounter a screen he didn't know what to do with, look around for instructions, find none, and close the browser. A single index card, written by hand, placed under the keyboard, that says:
1. Don't paste in anything you wouldn't write on a postcard. 2. For medical, legal, or money questions, double-check with the real expert.
— is more effective than any conversation you can have on Christmas Day. The card is the conversation, frozen and reusable.
How to set it up so she can't accidentally upgrade to the $20/month version
This is the part most people get wrong, and the one that produces the angry phone call three months later.
ChatGPT has a free version and a paid version. The free version is called ChatGPT Free and costs nothing. The paid version is called ChatGPT Plus and costs $20 a month. The paid version is genuinely useful for some people — it gives longer answers, image generation, and a few other features — but the parent who is reading her first email draft does not need it, and the cheapest way to set her up safely is to make sure she stays on the free tier until she or you decide otherwise.
The risk of accidental upgrade is real because the sign-up flow contains, at several points, screens that offer to "upgrade for unlimited" or "try Plus free for seven days." The seven-day trial is the trap. After seven days it auto-renews at $20 a month, and the charge appears on whatever card was used to create the trial, which — if your mother typed in her own card during a moment of confusion — will be her card. The Federal Trade Commission has named this specific pattern in enforcement actions across the subscription industry. It is not unique to OpenAI. It is unique to the era.
There are three concrete things to do during setup to close this off. First, sign her up using a Gmail or email login that does not have a payment method already linked. The free version does not require a card. If no card is on file, no upgrade can be silently triggered. Second, on the screen where ChatGPT offers Plus or a trial, click "Maybe later" or "Continue with free." That link is sometimes printed in pale gray. Look for it. It is always there. Third, after setup, go into Settings → Subscription and confirm the plan reads "Free." OpenAI's billing documentation at help.openai.com walks through the same screen if you want a reference. If you ever do want to upgrade her later — because she has started using image generation, say, or because the free-tier message limits are getting in her way — you can do it in about three clicks. Doing it on purpose is fine. Doing it by accident is the failure mode you are preventing.
One more piece of housekeeping: turn off the option in Settings → Data Controls that lets OpenAI use her conversations to train future models. This is on by default for free accounts and off by default for paid ones, which is the opposite of what most older users would expect if asked. The toggle is one click. It does not affect how the product works for her. It just keeps her drafts out of future training data, which is the version of "private" most people mean when they ask the question.
If you have read this far and the answer is yes, I want to set her up but I don't want to be the one on the phone with her about it three weeks from now, the binder I make is built for exactly that handoff. It is a printed, plain-English setup binder for ChatGPT — about thirty pages, with the two rules above already on the inside cover and the upgrade-prevention steps already done. It is $19 on the pricing page. A one-page free version of the first-day plan, if you would rather send that, is at the start link. And if she has already signed up and is asking the same safety question for herself rather than from a gift-giver's angle, I wrote that piece for her.
— Shawn Ivie Founder, Plain English Company