Can You Trust What AI Tells You? A Plain Guide for Seniors

AI tools like ChatGPT and Google Gemini can be genuinely useful. They explain things in plain words, help you write a letter, and answer questions in seconds. But you may have heard that they sometimes get things wrong, and that is true. The tricky part is that they get things wrong while sounding completely sure of themselves.

That does not mean you should avoid AI. It just means treating it like a clever, confident friend who is usually right but occasionally talks nonsense. This guide explains, in calm and plain English, when you can trust what AI tells you and when it pays to check. If you are helping an older parent get online, see our wider guide to helping a parent go online.

Why AI sometimes gets things wrong

An AI tool does not actually know facts the way a person does. It works by predicting what words are likely to come next, based on the huge amount of writing it learned from. Most of the time that produces a correct and sensible answer. But sometimes it fills a gap by inventing something that sounds right and simply is not. People who work with AI call this a “hallucination,” which is a polite word for confidently making things up.

It is not lying, and it is not trying to trick you. It has no idea it is wrong. That is exactly why a wrong answer can read just as smoothly as a right one. The mistakes can be small, like a slightly wrong date, or larger, like a phone number, a law, or a medical detail that does not exist.

The kinds of answers to be careful with

Some questions are riskier than others. Be extra careful, and always check elsewhere, when you ask AI about:

• Health and medicines. Doses, symptoms, and what to take. Studies have found AI gets a worryingly large share of health questions wrong. Ask your doctor, nurse, or pharmacist.
• Money and legal matters. Tax, wills, contracts, superannuation, entitlements. The details change often and the stakes are high.
• Phone numbers, addresses, and websites. AI can invent a number that looks real. Never ring a number it gives you for your bank or a government office. Find the real one yourself.
• Recent news and prices. AI may not know what happened this week, and it can guess at current prices that are out of date.
• Names, dates, and quotes. If it tells you who said something or exactly when, treat that as “probably” rather than “definitely.”

A simple habit: check before you trust

You do not need to be suspicious of everything. You just need one easy habit for anything that matters. Before you act on what AI tells you, ask yourself three quick questions.

1. Does this really matter? If it is a bit of trivia or a writing idea, relax. If it affects your health, money, or safety, check it.
2. Where would the real answer live? A bank detail belongs on your bank’s official site or statement. A medical answer belongs with a health professional. A government matter belongs on a .gov.au website.
3. Can I confirm it in one more place? A quick look at an official website, or a word with someone you trust, is usually all it takes.

A handy trick: you can ask the AI itself, “Are you sure? Where does that come from?” It will often add a useful caution, or admit it is not certain. That is a good reminder not to take the first answer as gospel.

When AI is perfectly safe to rely on

It would be a shame to avoid AI out of worry, because for plenty of everyday jobs it is both safe and genuinely helpful. You can happily lean on it to explain a confusing term, tidy up the wording of an email, suggest a recipe from what is in the fridge, draft a birthday message, or talk you through how a feature on your phone works. In these cases there is no harm done if it is a little off, because you are the one deciding what to keep.

If you would like a gentle introduction first, our guides on what AI actually is and what ChatGPT is are a calm place to start.

Asking in a way that gets better answers

You can lower the chance of a wrong answer by how you ask. A few simple habits help:

• Add “in Australia” when the answer depends on where you live, such as rules, services, or shops.
• Ask it to keep things simple, for example “explain this like I am new to it.”
• For anything current, ask it to search the web if it can, or check the date of its information.
• If an answer surprises you or seems too good to be true, it is worth a second look.

One last point worth knowing: scammers now use AI too, to write more convincing emails and even to copy voices. Trusting AI is one thing. Trusting a message just because it is well written is another. Our guide on whether AI is safe and how it affects scams covers that side in more detail.