Your First Conversation with an AI
Pick an assistant, set it up safely, and have a real conversation — asking follow-ups, asking it to explain simply, even switching to Bahasa Melayu. Plus what never to share, and what to do when it gets things wrong.
The first time most people use an AI assistant, they type something like “hello” or “what is the weather?” and feel a little underwhelmed. The second time, they ask it to explain the difference between takaful and conventional insurance in simple words, then ask a follow-up, then ask it to put the answer in a table — and suddenly they understand why so many people use it every day. This page is about getting you to that second conversation quickly, and safely.
Choosing an assistant
A chatbot assistant is an app or website where you type (or speak) to a large language model. Several big companies offer one, and all of the well-known ones have a free tier that is perfectly good for learning. Here are the main options, side by side and in no particular order As of 2026. Features, model names and limits change often, so treat the details as a starting point.
GPTChatGPTby OpenAI
The assistant that started the boom when it launched in November 2022, and still the most widely used. Website and apps for phones and computers; free tier with usage limits, plus paid plans. Can search the web, read uploaded files and images, and talk by voice.
CLClaudeby Anthropic
Known for careful, well-written answers and for handling long documents. Website and apps; free tier with usage limits, plus paid plans. Note that Anthropic requires users of its consumer app to be adults.
GEMGeminiby Google
Google’s assistant, available on the web, as an app, and built into many Android phones and Google services. Uses your Google account. Free tier, plus paid plans bundled with extra Google storage.
CPMicrosoft Copilotby Microsoft
Available on the web, as an app, in Windows and the Edge browser, and — under paid or workplace plans — inside Word, Excel and Outlook. If your employer provides Copilot, use the workplace version under your company’s rules.
METAMeta AIby Meta
Built into WhatsApp, Instagram, Facebook and Messenger in many countries, and available as a separate app. Availability and features vary by country, so check whether it appears in your apps.
+OthersMany more to explore
Other well-known options include DeepSeek, Perplexity (built around web search with sources), Grok (from xAI) and Le Chat (from Mistral). New ones appear regularly. Whichever you choose, download it only from the official website or your phone’s official app store; fake look-alike “AI” apps are a common scam.
For learning, the differences between the major assistants matter much less than how you use them. Pick one, get comfortable, and try a second one later to compare. There is no need to pay anything to complete this track.
Setting up safely
Most assistants let you sign up with an email address or an existing Google, Microsoft or Apple account. A few sensible habits:
- Use the official site or app. Type the address yourself or use the official app store. Do not follow links from unsolicited messages or adverts offering “free premium AI”.
- Protect the account. Use a strong password you do not use anywhere else, and switch on two-step verification if it is offered. Your chat history may contain things you would not want others to read.
- Check the age rules. Minimum ages differ between services — some allow teenagers with a parent’s permission, others are for adults only. Parents should read the terms before setting up an account for a child.
- At work, ask first. Many employers have rules about which AI tools are allowed and what may be put into them. Check before using any assistant for work tasks.
Look at the privacy settings before you chat
Before your first real conversation, spend two minutes in the app’s Settings, usually under headings like “Data controls”, “Privacy” or “Activity”. Policies differ between companies and change over time, so rather than trust anyone’s summary (including this one), look for yourself and answer these questions:
- Are my conversations used to train or improve the company’s models, and can I switch that off?
- Is my chat history saved, for how long, and can I delete it?
- Does the assistant have a “memory” of past chats, and can I see or clear what it remembers?
- Is there a temporary or incognito chat mode that is not saved to my history?
- Can human reviewers read conversations in some circumstances? (The privacy policy will say.)
A guided first conversation
Let us try a real conversation. Each prompt below has a copy button; paste it into the assistant you chose and read the reply before moving on. A prompt is simply whatever you type to the AI.
Step 1: ask something you are genuinely curious about
Notice that you told it who you are, what you want and how long. That small amount of context makes a big difference — the subject of Prompting Basics.
Step 2: ask a follow-up
Unlike a search engine, the assistant remembers the conversation so far (within its context window), so you do not need to repeat yourself.
Step 3: ask it to explain more simply
This is one of the most useful things an assistant can do: explain the same idea again and again, in different ways, without ever getting impatient.
Step 4: switch language
The major assistants work in many languages, including Bahasa Melayu, Chinese and Tamil, though quality is usually best in English. Try:
Read the answer critically. Is the Malay natural, or does it sound translated? Assistants sometimes mix in Indonesian words or overly formal phrasing — a good reminder that fluency varies by language.
Step 5: change the format
Regenerate, edit and talk
A few buttons and features make conversations much more useful. Their exact look varies between apps, but most assistants have them.
What does “regenerate” or “try again” do?
It asks the model to write a fresh answer to the same message. Because the model chooses words with some randomness (see temperature in What Is a Large Language Model?), you get a different version. Useful when the first answer misses the mark — and revealing: if two answers disagree on a fact, that fact needs checking.
Can I change what I asked?
Most apps let you edit a message you have sent (look for a pencil icon when you hover or long-press on it). The conversation then continues from your edited version. This is often better than piling on corrections.
When should I start a new chat?
Whenever you switch to an unrelated task. Long, wandering conversations can confuse the assistant, and older parts may drop out of its memory.
Can I just talk to it?
Yes. Most assistant apps offer a voice mode: tap the microphone or headphone icon and have a spoken conversation, which is handy while cooking or driving (hands-free only, please) As of 2026. You can also usually dictate a message and read the reply.
Can I show it a picture?
Many assistants can look at a photo you upload — a plant, a confusing appliance manual, a chart — and talk about it. Be careful what is in the picture: avoid images showing identity cards, documents with personal details or other people’s faces without their permission.
What not to share
Treat a chat with an AI assistant like a conversation in a public place with a helpful stranger who takes notes. Even with good privacy settings, your messages travel to and are stored by a company. Never type or upload:
Tip: you can usually get the same help by describing a situation in general terms. “How do I reply politely to a landlord who is late returning a deposit?” works just as well without names or addresses. The full list, privacy settings and how scammers use AI are on Staying Safe with AI.
When it is wrong
It will be, sometimes. Assistants can hallucinate — state false things confidently — and may not know recent events. Consensus No current assistant is reliable enough to trust blindly on facts that matter. Good habits:
- Push back. If something looks off, say so: “Are you sure? Please double-check that date.” Sometimes it corrects itself — but be aware it may also agree with you just to be agreeable, even when you are the one who is wrong.
- Ask for sources, then open them. If the assistant searched the web, it will often show links. A citation you have not clicked is not a citation.
- Check what matters elsewhere. Health, legal, financial or official information (tax deadlines, government aid rules, medicine doses) should be confirmed with the official source or a qualified person.
- Use the feedback buttons. A thumbs-down tells the company the answer was poor. (Check your privacy settings: sending feedback may share that conversation with the company for review.)
Before your first real conversation, what should you check in the app’s settings?
Data policies differ between companies and change over time, so look at the current settings yourself.
Two regenerated answers give different dates for the same historical event. What should you do?
Disagreement is a warning sign. Confidence and repetition are not evidence.
Which is safe to type into an assistant?
General requests are fine. Passwords and identity documents should never be shared.
// Pause and reflect
- After your first conversation, what surprised you — good or bad?
- What did you find in the privacy settings? Did you change anything?
- Name one task in your week where an assistant could genuinely save you time, and one where you would not use it.
Your notes are saved privately in this browser.
Ask your assistant about something you know very well — your hometown, your job, a hobby. Look for one thing it gets right that impresses you, and one thing it gets wrong or vague. Knowing where it fails on familiar ground is the best guide to how far to trust it on unfamiliar ground.
Key terms on this page
Sources & further reading
- OpenAI, “Introducing ChatGPT” (30 November 2022).
- Long Ouyang et al., “Training language models to follow instructions with human feedback” (2022).
- Official help and privacy pages of each assistant (OpenAI, Anthropic, Google, Microsoft, Meta) — read the current versions, as they change.
- Ethan Mollick, Co-Intelligence: Living and Working with AI (2024) — a practical guide to working alongside assistants.
- Personal Data Protection Act 2010 (Malaysia) — background on your rights over personal data.