Ever asked AI to make something, only for its answer to wander off in every direction? We want a simple recipe and get a complicated menu. We ask for help understanding a lesson and get an explanation packed with jargon. This is where effective AI prompting helps: it lets us explain our goals and needs more clearly.
A prompt is an instruction or input we give to AI. The following principles apply to ChatGPT, Gemini, Claude, and other services, although their capabilities and features differ. No magic phrase guarantees that the result will always be correct. These techniques are useful for studying, cooking, writing, planning trips, running community activities, and working. Let’s explore ten of them.
Start effective AI prompting with a clear goal
Before typing, decide what you need help with. Do we need ideas, a summary, a comparison, a draft, or a decision supported by analysis? “Good” and “interesting” need clear criteria for success.
For complex tasks, separate outlining, developing the content, and reviewing it. Correct the direction before AI produces a long piece of writing.
Example prompt: “Draft an email to a supplier asking for an updated delivery schedule. The goal is to get a confirmed date and an explanation of any delays. End by asking for a reply before Friday at 3 p.m. Western Indonesian Time (WIB).”
For studying, the goal might be understanding the material rather than getting a ready-made answer. For example: “Help me understand fractions through three progressively harder exercises. Don’t give me the answers immediately; offer hints after I try.”
Provide genuinely relevant context and materials
AI does not necessarily know our situation. Provide relevant materials, time limits, budget, equipment, skill level, and preferences. Relevant context is more useful than dumping every document into the conversation at once.
Separate the materials using labels such as “Context,” “Data,” and “Task.” Make it clear that the document’s contents are material to analyze, not new instructions. The Gemini prompt design guide also discusses context and specific instructions.
Example prompt: “I want to cook for two people in 30 minutes. I have eggs, tofu, carrots, rice, and one frying pan. Suggest two meals that aren’t spicy, use those ingredients, and explain the cooking sequence. If extra ingredients are needed, ask me first.”
Make sure attachments are readable. If AI cannot access a link, provide a relevant excerpt that is safe to share.
Define the audience and use a role when helpful
An explanation for a child, a first-year university student, a community organizer, and a company executive will naturally differ. Specify the reader’s knowledge level, their needs, the language, and the tone. Audience guidance helps AI choose its terminology and depth of discussion.
A role or point of view can help establish a lens for analysis. A tutor supports learning, an editor reviews a story, and an operations manager assesses implementation. However, asking AI to “be an expert” does not automatically make its answer accurate or replace a qualified professional.
Example prompt: “Explain photosynthesis to a 10-year-old. Take the approach of a patient tutor, use one simple analogy, and include three comprehension questions. Explain new terms, then describe the analogy’s limits so it doesn’t mislead.”
Specify the format and details you want
With effective AI prompting, specify the output format from the start. We can ask for a study schedule, shopping list, travel table, email, or presentation script. Define the length, sequence, columns, units, and details that must be included.
Avoid conflicting instructions, such as asking for a “very comprehensive” answer in just two sentences. Set priorities: a concise answer for an initial decision, followed by detail when needed. The right format makes the output useful for the next task.
Example prompt: “Create a seven-day study schedule for a mathematics exam. I have 45 minutes each afternoon. Use a table showing the day, topic, exercises, and break time. Prioritize fractions and percentages; leave the final day for revision.”
For a travel plan, add the budget limit, trip length, and participants’ needs. Ask AI to flag costs that have not been verified.
Explain what must stay and what to avoid
When editing, AI may overhaul parts we actually want to keep. Specify what must not change: figures, dates, direct quotations, terminology, links, a particular sequence, or the author’s point of view.
Pair restrictions with an alternative instruction. Instead of only saying “don’t sound stiff,” explain: “use short sentences and a polite form of address.” If constraints conflict, decide which takes priority.
Example prompt: “Polish this piece without changing figures, dates, product names, direct quotations, or URLs. Keep the inclusive ‘we.’ Don’t add personal experiences, testimonials, or new claims. If you find inconsistent figures, flag them for me to review.”
For a story, a constraint might be: “Improve the dialogue, but keep the characters, first-person viewpoint, and ending.” If you want new ideas, define how much creative freedom is allowed.
Give examples to make your standards clear
“Write in my style” becomes clearer when accompanied by writing samples. Choose one or two examples that represent the pattern, then explain what to follow: sentence rhythm, structure, level of formality, or approach to explaining things.
The Claude prompting guide includes examples among its techniques for guiding output. Examples show the standard we want; make sure AI does not also copy irrelevant facts or experiences.
Example prompt: “Style example: ‘It doesn’t have to be perfect right away. What matters is that we start and learn along the way.’ Follow its short sentences and warm tone to create three captions for community activities. Don’t copy the example or invent my experiences.”
If you provide an example of an unacceptable output, explain exactly why it was rejected.
Ask for supporting evidence, clarification, and uncertainty flags
A fluent answer is not necessarily correct. For effective AI prompting, specify which sources may be used and how missing information should be handled. Separate facts from assumptions, then ask for accessible references for important claims.
Example prompt: “Use only the report I have attached. Separate facts, interpretations, and unavailable data. Give the source page for every figure. If the period or unit is unclear, ask first. Don’t invent references.”
For current information, make sure the service actually has browsing access. Simply typing “use the latest data” does not provide that access. For travel, for example, ask for official sources for opening hours and ticket prices, then check again before leaving. This approach fits the habit of checking information discussed in my article about the spread of hoaxes.
Improve the result with specific feedback
Treat the first answer as a draft. If it falls short, identify what is wrong and what needs to change. “Not good enough” makes AI guess again; “the opening is too long” gives it something concrete to work on.
Example prompt: “The story is on the right track. Shorten the opening to two sentences, make the dialogue more natural, and reduce repetition. Keep the characters’ names, central conflict, and ending. Show the complete revised version.”
OpenAI’s prompting documentation discusses instructions, context, and examples for guiding output. In practice, keep a record of results that meet your needs and use them as the next benchmark. Improving the process also relates to becoming an effective and reliable worker.
For a recipe, feedback might be: “There are too many steps; simplify it for a beginner.” If the conversation is long, summarize the constraints before continuing.
Use skills for recurring tasks
Study notes, regular captions, reading summaries, and office reports can follow reusable rules. A skill can store instructions, workflows, examples, and reference materials. Anthropic’s explanation of Agent Skills describes this approach for specialized tasks.
Example prompt: “Use the catatan-belajar skill for the following material. Create a summary, five key terms, and a quiz without showing the answers yet. Preserve the source definitions.” In an office setting, a similar skill could summarize meeting decisions and follow-up actions.
This example assumes the skill is already available and readable. Skill support varies between products; if it is unavailable, use a prompt template or saved instructions. Skills support effective AI prompting, but still need testing and updating.
For more detail, read my article, 8 Tips for Creating AI Skills for More Consistent Recurring Tasks. It focuses on building, testing, and maintaining skills.
Protect your data and finish with a human touch
Use AI wisely. Don’t casually upload passwords, one-time passwords (OTPs), identity numbers, home addresses, children’s data, health information, or personal financial documents. At work, protect contracts, business strategies, and confidential company data; use services approved by your organization. Check company policies, service settings, user access, and data handling before uploading documents.
Example prompt: “Help categorize expenses using the following dummy data. Don’t ask for bank account numbers, addresses, or family identities. Flag the parts I need to fill in myself.” For workplace analysis, use anonymized examples in line with company policy.
Changing names may not be enough if other details still identify a person or company. The habit of securing our digital accounts remains important, too.
Finally, check and recheck facts, sources, calculations, logic, and tone. An additional AI review can help, but it is not independent verification. People provide the finishing touch and remain responsible for how the output is used.
Conclusion: give clear guidance, then use our judgment
Effective AI prompting starts with a clear goal, enough context, and standards we can check. Not every task needs a long prompt. Choose the relevant elements, provide examples when useful, and refine your instructions based on actual results.
AI can help with everyday activities and work. However, deciding whether a result is accurate, appropriate, and fit for use still requires human judgment. Let’s start with one simple need, set out the instructions, and check the result carefully before using it.
A short template we can use
“Task: [the output needed]. Goal: [what the output will be used for]. Context: [relevant situation and materials]. Audience: [who will read it]. Role: [a viewpoint, if needed]. Format: [type, length, and sequence]. Must include: [important details]. Preserve: [elements that must not change]. Avoid: [specific restrictions]. Example: [the standard to follow]. Sources: [permitted materials]. If important information is missing, ask before drawing conclusions. Flag assumptions and anything that cannot yet be verified.”
Fill in the sections you need. For simple tasks, the goal, materials, and format may be enough. For more complex needs, add context and constraints. Before sending, ask yourself: could someone reading these instructions for the first time understand them without guessing? After receiving the result, compare it with the original goal.
Give it a try, and keep thinking critically.
This post is also available in:
