Start here
New to AI? Start here.
Five short steps for a first-timer: what AI is, how to try it safely, how to ask, and how to check what it says. No jargon, and nothing to buy.
Checked on Oct 1, 2026; guides linked below checked September 24, 2026
1. What it is, in one minute
The chatbots people use today, such as ChatGPT, Claude, Gemini and Grok, run on large language models. They learn from a huge amount of text and practice predicting the next word, so they are very good at producing text that sounds right.2 Sounding right and being right are not the same thing. The full explanation is on How AI works.
Good to know
You do not need an account to read about AI, and you do not need to pay to start. Most of the big assistants have a free version. Prices change, so check Which AI?.
2. What it is good at, and where it fails
Good for
First drafts, summaries, rewriting an email, brainstorming, explaining a topic in simpler words, and organizing your own notes.
Be careful with
Facts, names, numbers, quotes, court cases, medical or legal questions, and anything recent. It can state false things confidently.1,2
Real example
In 2023 two New York lawyers were fined $5,000 after filing court cases that ChatGPT had made up.3
3. Start safely
- Do not paste secrets. No passwords, bank or card numbers, Social Security numbers, medical records, or work documents you are not allowed to share.
- Know the training setting. OpenAI says it may use content from consumer ChatGPT to train its models, with a setting to turn that off.4 Other bots have their own rules. Our Chat privacy guide lists the exact setting for each one.
- Be wary of voices and videos. Scammers can fake a relative’s voice. Agree on a family secret word. See Scams.8
4. Ask clearly
All three big makers give the same advice in their own guides: say what you want, give the background, say how you want the answer, and show an example if you can.5,6,7 Anthropic suggests a test: show your prompt to a colleague with no context and ask whether they would be confused.5
More rewrites, with sources, are on Prompting.
5. Check what it tells you
- Ask it for its source, then open the source yourself.
- For money, health, law, names and numbers, confirm with an official site.
- If it cites something you cannot find, assume it may not exist. See the lawyers example above.3
Use AI like a sharp intern: great for a first draft, a summary or ideas. You still sign off on the result.
Where to go next
Which AI?
Prices and strengths of the big seven.
Prompting
Before and after rewrites.
Glossary
Plain-English definitions.
Sources for this page
- Artificial Intelligence Risk Management Framework: Generative AI Profile (NIST AI 600-1). National Institute of Standards and Technology.
- Why language models hallucinate. OpenAI.
- Mata v. Avianca, Inc., No. 22-cv-1461 (PKC), Opinion and Order on Sanctions. U.S. District Court, S.D.N.Y. (via CourtListener).
- How your data is used to improve model performance. OpenAI Help Center.
- Be clear and direct (Claude prompting best practices). Anthropic.
- Prompt engineering. OpenAI.
- Prompt design strategies. Google (Gemini API docs).
- Scammers use AI to enhance their family emergency schemes. Federal Trade Commission, Consumer Advice.