Why this conversation matters

Older adults increasingly meet AI whether they choose to or not — in customer service chats, in news stories, and unfortunately in scams that use cloned voices and fake videos. A calm, jargon-free explanation from someone they trust helps them use the genuinely helpful parts and spot the dangerous ones.

Step 1: Start with what they already use

AI feels less alien when it turns out to be familiar. Point out the places it has quietly been helping for years:

  • Phone keyboards that suggest the next word.
  • Email spam filters that keep junk out of the inbox.
  • Map apps that predict traffic and suggest a faster route.
  • Voice assistants that set timers or play music on request.
  • Streaming services that suggest a programme “because you watched…”.

Then explain the new part: “The recent AI tools you’re hearing about do the same kind of pattern-spotting, but with language. You can ask them questions in normal words and they answer in normal words.”

Step 2: Use one simple analogy

Pick whichever fits the person best — and stick to one:

  • The well-read assistant: “Imagine someone who has read millions of books and websites and can answer almost anything instantly — but who never says ‘I don’t know’, so sometimes confidently makes things up.”
  • Super-powered autocomplete: “You know how your phone guesses the next word? This guesses the next word thousands of times in a row, so well that it writes whole answers.”
  • The new intern: “Quick, polite and helpful, but you’d still check their work before sending it to anyone important.”

Avoid words like “neural network”, “model” and “algorithm” unless they ask. If they want the fuller picture later, our plain-English guide How Do LLMs Work? is a good next read.

Step 3: A five-minute hands-on demo

Seeing beats hearing. Sit beside them, open a chatbot on a tablet or laptop with large text, and let *them* type or speak the questions. Good first tries:

  1. “Explain in simple words what a mortgage offset account is.” (Or any term that has confused them recently.)
  2. “Write a warm birthday message for my granddaughter who loves horses.”
  3. “Give me a simple recipe using chicken, rice and spinach.”
  4. “What questions should I ask my doctor about knee replacement surgery?”
  5. Take a photo of a plant in the garden and ask what it is.

Then make the most important point by showing it: ask about something they know far better than the AI — their hometown, their old trade, a family recipe — and look together for mistakes. Finding one teaches “double-check it” more effectively than any warning.

Step 4: Answer the questions they’ll actually ask

“Is it alive? Does it understand me?”

“No. It doesn’t think or feel. It’s very good at predicting what words should come next, which makes it sound like it understands.”

“Is it listening to me all the time?”

“A chatbot only sees what you type or say to it. But whatever you do tell it may be stored by the company, so treat it like a postcard, not a sealed letter.”

“Will it take everyone’s jobs?”

“It will change many jobs, as computers and the internet did. Right now it’s mostly used as a helper — for drafting, summarising and answering questions — with people checking the results.”

“Can I trust what it tells me?”

“For everyday things it’s often right, but not always. For anything about health, money or the law, use it to prepare questions — then ask your doctor, bank or a professional.”

Step 5: The safety talk (don’t skip this)

AI has made some old scams much more convincing. Keep this part short, calm and practical:

  • Voices can be copied. A phone call can sound exactly like a grandchild in trouble. If anyone asks for money urgently, hang up and call them back on the number you already have.
  • Agree a family code word that a real relative would know and a scammer wouldn’t.
  • Videos and photos can be faked, including of celebrities “recommending” investments. No genuine celebrity is personally offering them a deal.
  • Never type bank details, passwords, ID numbers or detailed health information into a chatbot.
  • Banks and governments won’t pressure you to act within minutes. Pressure is the warning sign.

Our guide on how to spot a deepfake goes deeper if they — or you — want to learn the warning signs.

Tips for making it stick

  • Stick to one app rather than showing several.
  • Increase text size and turn on voice input if typing is a struggle.
  • Write the steps down on paper: how to open the app, where to type, how to start a new chat.
  • Follow up a week later and ask what they’ve used it for.
  • Let them lead. Some people won’t be interested, and that’s fine — the safety talk is the part that really matters.