AI: the complete guide to using it well and not being fooled by it

Ofcom, the UK communications regulator, reported in April 2026 that just over half of UK adults (54%) now say they use applications such as ChatGPT, Copilot or Gemini. In the United States, a Pew Research Center survey in early 2025 found that 34% of adults had ever used ChatGPT, falling to 25% of people aged 50 to 64 and 10% of those 65 and over.

So some readers of this page use an AI assistant every day, and others have never opened one. This guide is written for both. It covers what the thing is, what it is good and bad at, and how the same technology is being turned against you, with a link to a fuller guide on each.

What an AI assistant actually is

An AI assistant, or chatbot, is a program you type or talk to in ordinary language. The main ones at the time of writing, in alphabetical order, are ChatGPT, Claude, Copilot and Gemini. They work in broadly the same way, and this site does not recommend one over another.

Underneath is a “large language model”. The UK’s National Cyber Security Centre (NCSC) describes it as an algorithm trained on a large amount of text, typically scraped from the open internet, which analyses the relationships between words and turns them into a probability model. In plain words: it has read an enormous amount of writing and produces the words most likely to come next.

That explains both its talent and its flaw. It writes fluently on almost anything. It does not look facts up in the way a reference book does, so, as the NCSC puts it, it can get things wrong and “hallucinate” incorrect facts. The makers say the same. OpenAI’s help page warns that ChatGPT “might sound confident—even when it’s wrong”. Fluent and wrong at the same time is the thing to remember.

What it is good for, and what it is not

The strengths are all about words and ideas:

• Drafting a letter, an email or a complaint for you to edit.
• Explaining a confusing document in plain English.
• Planning a trip, a menu or a project and listing what you have forgotten.
• Practising a language (see learning a language after 50).
• Summarising something long.
• Brainstorming names, gifts, routes, opening lines.
• Getting your questions in order before an appointment.

The weaknesses are just as consistent:

• Facts stated confidently but wrongly. OpenAI lists incorrect dates and definitions, and fabricated quotes, studies and citations.
• Anything that depends on your circumstances. It does not know your health, your finances or your local rules unless you tell it, and there are things you should not tell it.
• Up-to-the-minute information. Anthropic’s help page says Claude may struggle with current events because of the limits of its training data. Assistants that search the web do better, but Anthropic still advises reviewing the sources they cite.

OpenAI’s own summary is the best one-line rule: “Use ChatGPT as a first draft, not a final source.”

Your first conversation

You cannot break it, and you cannot offend it. Type what you want as you would say it to a helpful stranger, then say what was wrong with the answer and ask again. Most of the skill is in that second message.

A step-by-step first session, with things to try, is in your first conversation with an AI assistant.

Checking what it says

The habit that separates a good user from a fooled one is checking the things that matter. OpenAI’s advice: “Always verify quotes, data, technical information or references to external documents.” Anthropic says not to rely on Claude as “a singular source of truth”.

This site is a working example. It is researched and written with the help of AI, as the footer says, and every fact is checked against the sources listed with each article: a helper whose work gets checked. How to do that quickly yourself is covered in when AI gets it wrong: how to check.

Privacy: what not to type in

What you type does not vanish. The NCSC’s advice is not to include sensitive information in queries to public chatbots, and not to submit anything that would cause problems if it were made public. Google’s privacy notice for Gemini is franker still, asking users not to enter “confidential information that you wouldn’t want a reviewer to see or Google to use to improve our services”.

So keep out passwords, bank and card details, government ID numbers and other people’s private business. Remove names and account numbers before pasting in a letter. The settings worth knowing are in what to tell an AI and what to keep private.

Health questions

An assistant can be useful before an appointment: explaining a term in plain language, or helping you turn a muddle of worries into three clear questions. It cannot examine you and does not know your history. Google tells Gemini users directly: “Don’t rely on Gemini Apps for diagnosis, treatment, medical advice, or legal, financial, or other professional help.”

Everything here is general information, not medical advice. Check with your GP (family doctor) or pharmacist before acting on anything a chatbot says about your health. The sensible way to use it is in using AI for health questions before an appointment.

The other side: how criminals use it

The technology that drafts your letter will also draft a fraudster’s.

Cloned voices. The US Federal Trade Commission (FTC) warns that a scammer needs only a short audio clip of a family member’s voice, which could come from content posted online, and a voice-cloning program. The call that follows sounds like someone you love, in trouble, needing money now. See AI voice-clone and family-emergency scams.

Deepfake videos. Which?, the UK consumer group, reported in 2024 that fake videos of well-known people, with Martin Lewis and Elon Musk among the most impersonated, were circulating on Facebook and Instagram, and that the majority were linked to cryptocurrency investments. A famous face recommending an investment in an advert is not evidence of anything. This is general information, not financial advice.

Fake reviews and profiles. The FBI says criminals use AI to create large numbers of fictitious social media profiles, and the FTC’s 2024 rule on fake reviews specifically names AI-generated ones. How to spot AI fakes: photos, videos and reviews covers the tells.

The good news is that the old defence still works, because the scam still depends on hurrying you. The FTC’s advice on a cloned voice is: “Don’t trust the voice.” Call the person back on a number you know is theirs. The FBI adds two more: agree a secret word or phrase with your family, and if a bank or organisation calls, hang up, find its number yourself and ring that. The pressures underneath have not changed, and they are set out in how to spot a scam text, email or call, along with where to report in the UK and US.

Where to get free help learning

UK. Age UK’s Digital Champion Programme, run with 40 local Age UKs, offers one-to-one support and IT drop-in sessions, and Age UK also describes loan devices for people without their own; use “Find services near you” on its website to see what your local Age UK provides. Public libraries in England and Wales provide free internet access and community activities, according to Gov.uk; libraries elsewhere in the UK are worth asking too. The Open University’s OpenLearn site has a free short course, Understanding Generative AI. And u3a, with more than 1,000 locally run groups, describes itself as a movement whose members come together to learn for fun in retirement.

US. Senior Planet, run by OATS, a nonprofit affiliated with AARP, offers free online classes in technology and other subjects, with no age restriction and no AARP membership needed; in 2021 AARP put the number at about 50 a week. Your public library is also worth asking.

Your first week

• Monday: Open one assistant and ask it about a subject you know well. See where it is right, and where it is not.
• Tuesday: Ask it to draft a short letter or email you have been putting off. Edit it until it sounds like you.
• Wednesday: Paste in a confusing paragraph, personal details removed, and ask for plain English.
• Thursday: Take one fact it gave you this week and check it against an official source.
• Friday: Find the assistant’s privacy settings and read what it keeps.
• Saturday: Agree a safe word with your family, and tell them why.
• Sunday: Write down three questions for your next appointment with its help, or plan a day out.

Seven small steps. By the end you will know what the fuss is about, what it is useful for in your own life, and why a confident voice, human or otherwise, is worth one phone call to check.