Skip to main content

How to ask an AI chatbot your first question and get an answer worth using

An AI chatbot answers questions you type in ordinary words. Say exactly what you want, and check anything important, because the makers warn the answers can be wrong.

KBY Everyday Tech 9 min read Checked 22 November 2026 ChatGPT, Gemini, Copilot and Claude on a phone or computer, as their makers describe them on 22 November 2026
An older woman's hands typing on a laptop keyboard, with handwritten notes on the table beside it
Photo: RDNE Stock project on Pexels

Pick one chatbot. Open its website or app. Type a question as if you were asking a well-read friend.

Read the reply with a pinch of salt. That is the whole of getting started.

ChatGPT (made by OpenAI), Gemini (Google), Copilot (Microsoft) and Claude (Anthropic) all work this way. None of them asks you to pay before your first question.

You don't need to try all four. Starting with one linked to an account you already have is a sensible place to begin. That might be a Google account for Gemini, say, or a Microsoft account for Copilot.

Nothing here ranks them.

What a chatbot is, in plain words

A chatbot is a program you hold a written conversation with. You type a message, which the makers call a prompt. It types a reply.

You can then reply to its reply. It keeps track of what you said earlier in the same conversation.

The ones in this guide are built on what the industry calls a large language model. Google's page Learn about generative AI explains it simply.

The model was trained on a huge amount of text. It writes by predicting which words are likely to come next. It bases this on your prompt and what it has written so far.

Google is clear that this kind of AI "is not a human being. It can't think for itself or feel emotions."

That explains both what chatbots are good at and where they go wrong. They are quick at drafting, rewording, summarising and explaining.

But predicting likely words is not the same as looking up facts. So a chatbot can write something that sounds right and isn't.

The makers have a name for this, a hallucination. There is a section on it below.

What each maker says you need

Each of the four works in a web browser and as an app for iPhone and Android phones. A web browser is the program you use to visit websites, such as Chrome, Safari or Edge.

An account is a sign-in, usually with your email address. It lets the service save your past conversations.

ChatGPT

OpenAI's help page on the ChatGPT home page says ChatGPT is free to use. It says you can try it at chatgpt.com without creating an account in supported countries, "including in Europe and the UK".

Without an account you can have only one conversation at a time, and it isn't saved. Creating an account lets you keep your chat history. OpenAI recommends a personal email address for it, not a work one your employer controls.

The ChatGPT Free Tier FAQ says free users have unlimited everyday text chats. It says things like file uploads and image creation have separate limits. It also says ads may appear in the free tier in certain countries.

OpenAI's mobile app guide says the app is on the App Store and Google Play for iPhone, iPad and Android.

Gemini

Google's page What you need to sign in to Gemini Apps says you can use some features of the website, gemini.google.com, without signing in. To use more features or keep your history, you sign in with a personal Google account.

Google lists the United Kingdom among the places the website works. It says you need to be 13 or over (or the applicable age in your country) to use Gemini with a personal account.

When you're signed out, Google says you can generally ask questions and get simple text answers. But you lose your past chats, image creation and file uploads. On an Android phone, the Gemini app needs you to be signed in.

Google's mobile app guide says there is a Gemini app for iPhone as well as Android. On some Android phones, switching to Gemini replaces Google Assistant as the phone's assistant. Read the screen before you tap yes.

Google sells paid Google AI plans as an upgrade for "expanded access to features and models". That is according to its page on Gemini limits and upgrades.

Copilot

Microsoft's page Get started with the Microsoft Copilot app says where to find it.

  • The web version is at www.copilot.com.
  • There are apps for Windows and Mac.
  • The phone app is in the Apple App Store (iPhone with iOS 16.0 or later) and the Google Play Store.

Microsoft describes a personal version for individuals, "with or without a Microsoft 365 Personal, Family, or Premium subscription".

You can sign in with a personal Microsoft account, an Apple ID or a Google account. Microsoft says a personal Microsoft account is the kind used for Outlook, OneDrive and Xbox.

Microsoft says signing in "unlocks more, including your chat history, more image creation, longer conversations, extended voice sessions, and other features".

Claude

Anthropic's Get started with Claude page says you can use Claude in three places.

  • In a web browser at claude.ai.
  • In a desktop app for Mac or Windows.
  • In the app for iPhone or Android.

You must be at least 18 and in one of Anthropic's supported locations. The free plan has a usage limit that resets every five hours. Anthropic says the number of messages you can send varies with demand.

There is no password to remember. Anthropic's page on logging in says you either choose "Continue with Google" or enter your email address. It then emails you a link to sign in.

Writing a first question that gets a useful reply

The makers' advice for beginners is much the same. Anthropic says to talk to Claude "like you would a coworker or friend". It also says to start simple, be specific, and refine your request or ask follow-up questions.

OpenAI's article How do I create a good prompt for an AI model? gives the reason. If you asked a person to "do this task for me" with no detail, they would probably not do what you wanted. The same goes for a chatbot.

Put together, the advice comes down to four habits.

  1. Say exactly what you want. "Write a message to my neighbour" is vague. "Write a short, friendly text to my neighbour asking if they could take in a parcel for me on Thursday" tells it the length, the tone and the point.
  2. Give it the background. OpenAI says to identify the task, provide the context it needs, and set the tone you want. Who is it for? What do you already know? What has already happened?
  3. Keep the conversation going. The first answer is a starting point. OpenAI calls this working "iteratively". Read the reply, then ask for changes, such as "shorter", "less formal", "use British spellings" or "give me three other options". Google's page on using Gemini says you can also edit a prompt you've already sent. Gemini then writes a new reply.
  4. Ask it to explain simply. If a reply is full of jargon, say so. OpenAI's prompt advice for ChatGPT suggests describing the tone you want in words like "friendly" or "formal". You could type: "Explain that again in plain English, as if I've never heard of it."

Everyday things to try first

These are ordinary jobs that suit a chatbot. The result is easy for you to read and judge before you use it.

Tidy up an email. Paste in a draft and add: "Make this email to my landlord about the broken boiler clearer and more polite, but keep it firm. Keep it under 150 words."

Read what comes back. Change anything that isn't true, and send it yourself.

Make sense of a confusing letter. Type out the part you don't understand. Then ask: "Explain what this paragraph means in plain English, and tell me what it is asking me to do."

Leave out your name, address, account numbers and anything else that identifies you. There is more on that below.

Treat the answer as help understanding the letter, not advice. On money, health or the law, the sender or a qualified adviser has the final say. Google says plainly not to rely on Gemini's answers "as medical, legal, financial or other professional advice".

Plan the week's meals. "Plan five quick evening meals for two adults, one of whom is vegetarian, using things from a normal supermarket. Then give me one shopping list."

If you don't like a dish, say which one and ask for a swap.

Plan a day out or a trip. "I'm visiting York for a weekend in March with my mum, who can't walk far. Suggest a plan for each day with places close together."

Google's own examples include asking Gemini to draft a packing list for a weekend away. Check opening times, prices and train times on the official websites before you book anything. A chatbot can be out of date.

When the chatbot gets it wrong

All four makers say, in their own help pages, that their chatbot can be wrong.

  • OpenAI's article Does ChatGPT tell the truth? says ChatGPT can produce incorrect or misleading answers. It says ChatGPT can sound confident even when it is wrong, including made-up quotes and references to sources that don't exist. Its advice is to use ChatGPT "as a first draft, not a final source" and to check important information.
  • Google's page Learn about responses from Gemini Apps says Gemini "may sometimes produce inaccurate, offensive or inappropriate information". Its page on generative AI adds that it can misunderstand. Ask about bats, the animals, and you might get cricket bats.
  • Microsoft's Transparency Note for Microsoft Copilot says Copilot can produce content "that might sound reasonable but is factually inaccurate". It encourages people to double-check facts before making decisions.
  • Anthropic's article Claude is providing incorrect or misleading responses says Claude can write "things that might look correct but are very mistaken". It says Claude should not be relied on "as a singular source of truth".

So check before you act. When a reply lists its sources as links, open them and read the original. Anthropic and OpenAI both suggest this.

For anything about your health, money, legal rights, benefits or tax, go to the official source. That could be the NHS website, GOV.UK or the organisation involved.

If a chatbot says it has done something outside the conversation, such as sending an email, check that it really happened. Anthropic's article on links that don't work and false claims says Claude can "hallucinate its capabilities". It says Claude can claim to have sent an email when it has no access to email at all.

Recent events are another weak spot. OpenAI says its models only know about events after their training date when a tool such as web search is used. Anthropic says Claude can get confused about current events.

What to keep out of the chat

What you type is sent to the company that runs the chatbot. In its blog post ChatGPT and large language models: what's the risk?, the UK's National Cyber Security Centre says your questions are visible to the provider.

The centre recommends not including sensitive information in them. It also recommends not asking anything that would cause you problems if it were made public.

That means no passwords, bank details, National Insurance number or medical records. It also means nothing private about other people.

The makers also give you some control over what happens to your conversations. For example:

  • OpenAI's help page says you can turn off "Improve the model for everyone" under Data controls in ChatGPT's settings.
  • Google offers a temporary chat in Gemini that doesn't appear in your activity or your recent chats.

Our guide to using AI chatbots safely covers what to keep private and those settings in more detail.

Your chatbot account is worth protecting like any other. Turning on two-step verification for the email account it is linked to is a good start.

Common questions

Do I have to pay to use an AI chatbot?
No, not to start. OpenAI says ChatGPT is free to use and Anthropic has a free Claude plan. Google lets you use some Gemini features without even signing in and sells paid Google AI plans as an upgrade, and Microsoft says its personal Copilot works with or without a Microsoft 365 subscription. Free use comes with limits, such as how much you can do in a few hours, and paid plans raise them.
Which AI chatbot should I use?
This guide does not rank them. All four do the everyday jobs described here. A practical way to choose is to start with the one linked to an account you already have: Gemini uses a Google account, Copilot a Microsoft account (or an Apple ID or Google account), and ChatGPT and Claude both let you sign up with an email address. Claude is for over-18s only, according to Anthropic.
Can I trust what an AI chatbot tells me?
Not without checking. OpenAI, Google, Microsoft and Anthropic all say in their own help pages that their chatbot can give wrong answers that sound confident. Use the reply as a first draft or a starting point, open any sources it links to, and check anything about health, money or the law with an official source such as the NHS website or GOV.UK.

Sources

Every setting, menu path and figure in this guide comes from these pages, read on 22 November 2026. Menus and prices change with updates: check the source before you rely on one.

  1. The ChatGPT home page (OpenAI)
  2. ChatGPT Free Tier FAQ (OpenAI)
  3. Using the ChatGPT mobile app (classic experience) (OpenAI)
  4. How do I create a good prompt for an AI model? (OpenAI)
  5. Prompt engineering best practices for ChatGPT (OpenAI)
  6. Does ChatGPT tell the truth? (OpenAI)
  7. What you need to sign in to Gemini Apps (Google)
  8. Use Gemini Apps (Google)
  9. Get started with the Gemini mobile app (Google)
  10. Where you can use the Gemini web app (Google)
  11. Gemini Apps limits and upgrades for Google AI subscribers (Google)
  12. Learn about responses from Gemini Apps (Google)
  13. Learn about generative AI (Google)
  14. Get started with the Microsoft Copilot app (Microsoft)
  15. Transparency Note for Microsoft Copilot (for individuals) (Microsoft)
  16. Get started with Claude (Anthropic)
  17. Log in to your Claude account (Anthropic)
  18. Claude is providing incorrect or misleading responses. What's going on? (Anthropic)
  19. Claude is producing links that don't work and falsely claiming that it has sent emails or produced external documents. What's going on? (Anthropic)
  20. ChatGPT and large language models: what's the risk?, Blog post, published 14 March 2023 (NCSC)

Spotted something wrong or out of date? Tell us and we will correct it.

About this page

Dates

Published

Corrections

No corrections are listed for this page. Some review sweeps corrected pages without naming each one; those are in the corrections log.

Keep reading

More Everyday Tech guides

All Everyday Tech guides
Share this guideXLinkedInFacebookEmail

Was this useful?

Reports go to the site owner and are used to correct pages. Corrections are listed at /corrections.

All Everyday Tech guides

The KBY Brief

A short email about twice a week with new runbooks, configuration traps and fixes from KBY Technologies.