Ways to use ChatGPT for best results

ChatGPT reached an estimated 100 million users within two months of its launch in November 2022, according to a UBS analysis, and in September 2026 OpenAI said it had 1.2 billion weekly users. Yet many people still don't get the most out of the chatbot, mostly because of the way they interact with it. In this article we will show you how to get the best results with ChatGPT.

You can try ChatGPT without an account, although signing in lets you keep your chat history and use more features. All you need to do is type in your text and get the information you seek. But there are some basic principles to look into to get the right answers in a comprehensible form. When giving the chatbot an assignment, consider:

1. giving it context: explain who the answer is for, what you need it for, and the length and format you want (a table, a list, an email). OpenAI’s own prompting best practices recommend clear, specific requests with enough background
2. paraphrasing questions in forms that the chatbot will understand, especially when it's giving you wrong answers or refusing to provide one
3. describing the preferred complexity of the response you seek. For example, you can instruct the chatbot to explain a concept to a five-year-old or instruct it to add references and examples to the answers it provides
4. naming the industry niche you want the response to come from, to avoid a generalised answer that lacks the specific terminology you seek. For example, you can instruct the chatbot to give you replies like a medical doctor or a football analyst
5. instructing the chatbot to respond in a style or form for better results. For instance, you can instruct it to write a historical romance story or a contemporary romance story. It all depends on how creative you are; for more ideas, see five ways content creators can utilise generative AI
6. using custom instructions or projects for recurring tasks, so you don’t have to repeat the same background in every chat

Be aware that ChatGPT can still give incorrect information that sounds convincing, a weakness critics have pointed out since its early days (see this warning from a Google executive). Its built-in knowledge ends at a training cutoff date; it can search the web for recent events, but you should still check important facts and the sources it cites. OpenAI itself warns users that ChatGPT can make mistakes.

The fun part is that you can make the chatbot respond like a human with emotions. Although the chatbot doesn't have emotions, it's very good at using emoticons and emoji, and you can easily tell it to use them in its replies. More importantly, if a reply stops halfway or you want more detail, simply ask it to continue or to expand on a specific part; with long code, ask it to output the rest from the point where it stopped.