From Knowing to Mastering AI: 4 Prompting Habits You Can Start Using Today

Many people have already made it a habit to use AI chat tools daily for all kinds of tasks, such as organising notes, rewriting articles, planning itineraries, drafting emails, creating presentation outlines, and condensing information into key points. However, the effectiveness of AI often depends not on the tool itself, but on how you ask.

Prompting is less a technical skill than a communication ability. It is very similar to communicating with people: when a question is too vague, AI can only guess and keep asking for clarification; on the other hand, when you provide clear context, goals, and constraints, AI is more likely to produce content that can be used directly, while also reducing the time spent on revisions.

This article introduces 4 simple prompting habits to help you communicate your needs clearly in one go, reducing the cost of back-and-forth exchanges. A universal Prompt template is also provided at the end, which can be applied directly to any scenario you encounter in the future (recommended to save for reference).

Clearer questions lead to better AI responses

The following examples demonstrate how "information density" directly affects the quality of AI output.

Less ideal prompt:
"Help me summarise this article."

Without specifying the purpose, audience, format, or length requirements, AI may produce output that is too long, too short, or missing the key points you need, and you will still end up spending time reworking it.

More ideal prompt:
"Below is an article. Please summarise it into 6 key points in Cantonese, presented in bullet point format, with each point no longer than two sentences; add a summary of no more than 60 words at the end, in a neutral tone, suitable for reading before a meeting."

When you specify the purpose, format, length, and tone all at once, AI can usually deliver a usable version much faster.

Four Prompting Habits to Improve the Quality of AI Output

1.Give Context

AI does not automatically understand your situation, audience, or preferences. To improve the usability of its output, you need to clearly state the key information, such as what you need it for, who will read it, what materials you already have, and what style or references you have in mind.

Comparison example (writing an email)
Less ideal: "Write me an email."
More ideal: "I need to write a formal email to confirm that the meeting time has been moved to next Wednesday morning. The tone should be polite and concise."

Comparison example (rewriting an article)
Less ideal: "Help me fix this paragraph."
More ideal: "Below is a paragraph in Chinese. Please retain the original meaning and rewrite it in clearer, more fluent formal written Chinese; avoid exaggerated language; and list the three main areas of revision you have made."

When it comes to written tasks, "setting a role" can also be very effective. You can open with a single line such as "You are a social media copywriter," "You are a product manager," or "You are a teacher" to help AI quickly align with the appropriate tone, depth, and structure.

2.Set Constraints

Output without constraints tends to go off-topic, run too long, or be unsuitable for your intended use. A more effective approach is to clearly state both what you want and what you do not want.

• You can specify what you want: word count, paragraph structure, number of key points, whether examples are needed, language, and tone.
• You can also specify what you do not want: no jargon, no marketing-style language, no unverified figures or sources, no tangents unrelated to the topic.

Comparison example (writing a social media post)
Less ideal: "Write me a post."
More ideal: "Please write a Cantonese Instagram caption for the following event recap, within 150 words; the tone should be professional yet easy to read; avoid exaggerated language or excessive adjectives; and provide 3 alternative headlines at the end."

3. Ask AI to Show Its Work

AI sometimes gives conclusions directly, but the reasoning or judgment behind them may not always be correct. For tasks involving analysis, recommendations, or information synthesis, it is advisable to ask AI to clearly explain its reasoning and judgment criteria, so you can assess its reliability and make corrections quickly.

You can add simple instructions such as:
"Please explain your criteria for judgment before giving the conclusion."
"Please clearly state the risks, limitations, and uncertainties."
"Please explain the reasoning behind your analysis before making recommendations."

Example (replying to a message)
"I need to reply to someone: I want to politely decline an invitation while maintaining a good relationship. Please provide two versions (more direct / more tactful), each within 60 words. Please clearly state the risks, limitations, and uncertainties: for example, potential misunderstandings each version may cause, the types of relationships and situations each version is better suited for (close / not close, formal / informal), and what background information you still need me to provide (my relationship with the person, the nature of the invitation, the reason for declining) in order to write a more tailored response."

4.Always Verify

One of the common risks of AI is that it may present inaccurate information in a very confident tone. This is especially true when the content involves data, citations, legal or medical advice, policies, or specific product specifications, all of which require careful verification.

It is recommended to prioritise checking the following types of content:

Data and statistics: whether they are outdated, and whether the sources are reliable.
Citations and references: whether they actually exist and can be verified.
Links and resources: whether they are valid and whether the content matches.
Key details: whether definitions, terms, steps, and cause-and-effect relationships are accurate.

A safer approach is to treat AI as a tool for drafting and accelerating work, rather than as a final answer. Any content intended for public release, use as an official document, or that may influence others' decisions should be verified and revised by the user.

A Reusable Prompt Template (Recommended to Share and Save)

You can copy and share the following template, replacing the content in brackets as a standard format for your everyday prompting:

"You are (role). My goal is (objective), and the use case is (purpose / audience).
The background or materials are as follows: (paste content here).
Please output in (language) with a (tone) tone, in the format of (format), with a length of (word count / number of paragraphs).
Please avoid (content you do not want to appear).
Finally, please indicate which parts of the response are based on the materials I provided and which would require additional information to confirm; and raise three questions you need me to answer in order to make the response more accurate."

Conclusion: Make AI Work for You, Not the Other Way Around

Good prompting is, at its core, the ability to articulate your needs clearly and break down your problems effectively. When you can provide sufficient context, set clear constraints, ask for reasoning to be shown, and verify the output thoroughly, AI is far more likely to become a reliable asset, helping you complete tasks more efficiently while improving the usability and accuracy of the content produced.