Cart 0
Written by Jacob Sipos\ Published Sep 21, 2026
How To Write Better AI Prompts
A better prompt is not a secret code. It is a better assignment.
Picture yourself handing a task to a smart new assistant. If you say, “Make this better,” the assistant has to guess. If you explain the goal, audience, source material, limits, and finished result, the assistant has a much clearer target.
That is prompting. You are communicating what work needs to be done. Better prompts reduce guessing, make outputs easier to review, and often save the back-and-forth that happens when the AI has to infer what you meant.
Write an email.
I can, but I have to guess the audience, purpose, tone, details, and desired next step.
Goal: Follow up after yesterday’s sales call.\
Audience: Potential client.\
Context: They asked about implementation timing.\
Output: 150 words, professional and warm.\
Boundary: Do not invent pricing.
Now I have a clear assignment.
🎯 Goal
🧠 Context
📄 Output
🛡 Boundary
First, What Is an AI Prompt?
A prompt is the instruction or input you give an AI system. It can be a sentence, a long assignment, a file plus instructions, an image, or a conversation that develops over several messages.
OpenAI’s current guidance defines prompt engineering as designing and refining inputs to guide a model’s responses, while emphasizing clear, specific instructions, useful context, and iterative refinement. OpenAI Help
The Core Idea
The AI cannot automatically see the picture in your head. A good prompt transfers the important parts of that picture into the conversation.
You do not need to write a giant prompt every time. A simple question can be perfect for a simple task. Add detail when the task has more ways to go wrong.
Use Goal + Context + Source + Output + Boundary
A practical framework is to give the AI five things when they matter. OpenAI’s 2026 small-business prompting materials similarly teach Goal, Context, Output, and Boundary, while other OpenAI guidance emphasizes source material and examples. OpenAI Academy
GOALWhat job should the AI accomplish?
CONTEXTWho, why, background, situation.
SOURCEWhat information should it rely on?
OUTPUTFormat, length, tone, structure.
BOUNDARYWhat must it avoid, flag, or leave for review?
Goal: Draft a customer update about the delayed launch.\
Context: Customers expected release Friday. The new target is Tuesday because final testing is incomplete.\
Source: Use only the project notes below.\
Output: Write a 180-word email with a clear subject line. Sound calm, accountable, and professional.\
Boundary: Do not invent technical details or promise anything not stated in the notes.
You do not have to label every prompt exactly this way. The framework is a checklist for the information the model may need.
Context Is Often the Difference Between Generic and Useful
Vague
Give me ideas to increase sales.
The AI has to guess the business, customer, price, channel, problem, resources, and objective.
→
Grounded
We sell an online exam-prep membership to adults preparing for a professional certification. Most customers discover us through search and social media. Give me 10 ideas to improve trial-to-paid conversion without lowering the price.
Context can include the audience, company, product, previous decisions, constraints, deadline, skill level, geography, objective, or anything else that changes what a good answer looks like.
But more text is not automatically better. Current OpenAI model guidance recommends leaner prompts where possible and avoiding repeated instructions. Give the model high-signal information, not every thought you have ever had about the project. OpenAI model guidance
Tell the AI What the Finished Product Should Look Like
“Make this professional” leaves enormous room for interpretation. Describe the deliverable.
FormatEmail, memo, table, article, outline, checklist, script, FAQ, JSON, presentation structure.
Length100 words, one page, five bullets, 10 options, three paragraphs.
AudienceExecutives, customers, beginners, technicians, students, parents, investors.
ToneWarm, concise, formal, reassuring, persuasive, neutral, energetic.
OpenAI’s current prompt guidance recommends being specific about the desired context, outcome, length, format, and style, and notes that examples can help communicate the desired output format. OpenAI Help
When Accuracy Depends on Your Information, Give the AI the Information
Instead of asking AI to guess what your policy says, provide the policy. Instead of asking it to invent a project update, provide the project notes. Instead of asking it to summarize sales performance from memory, provide the current spreadsheet.
YOUR MATERIAL
CLEAR INSTRUCTION
AI TRANSFORMATION
HUMAN REVIEW
Think of the difference between “Write our refund policy” and “Using the attached approved refund policy, create a customer-friendly FAQ without changing the policy.”
The second prompt gives the AI an authoritative source and a transformation task. That reduces unnecessary invention.
If You Can Show What “Good” Looks Like, Do It
Sometimes explaining a style is harder than showing it. Give the AI one or two examples of the pattern you want.
Here is the style I want:\
\
Technical: “Authentication credentials are invalid.”\
Plain language: “The email or password does not match our records.”\
\
Technical: “Your subscription has exceeded the permitted seat allocation.”\
Plain language: “Your plan has reached its user limit.”\
\
Now rewrite the following message using the same plain-language approach: [message]
OpenAI’s prompt-engineering guidance explicitly recommends showing desired output formats through examples when useful. OpenAI Help
Tell the AI What to Do When Information Is Missing
One of the most useful prompt additions is not a style instruction. It is a rule for uncertainty.
Do not invent“If the source does not contain the answer, say that the information is not provided.”
Flag uncertainty“Separate confirmed facts from assumptions or recommendations.”
Ask first“If a missing detail would materially change the result, ask me before drafting.”
Keep human review“Mark any legal, financial, safety, or policy claim that should be verified before publication.”
OpenAI’s current small-business framework specifically includes a Boundary component, such as flagging missing information, keeping something as a draft, or not inventing details. OpenAI Academy
The First Prompt Does Not Need to Do Everything
Prompting works well as a conversation. OpenAI’s current ChatGPT guidance recommends iterative refinement: review the first response, then adjust wording, context, or requirements based on what is missing. OpenAI Help
ASK
REVIEW
CORRECT
REFINE
Useful follow-ups include: “Make paragraph two easier to understand.” “Keep the structure but reduce the length by 25%.” “Give me three alternatives for the opening.” “You misunderstood the audience. This is for first-time customers.” “Before revising, tell me what information is missing.”
Specific feedback is much more useful than repeatedly saying, “Make it better.”
For Complex Work, Define the Stages
Large tasks can often be improved by creating checkpoints rather than demanding a finished answer immediately.
Stage 1: Understand.
Ask the AI to summarize the assignment and identify missing information.
Stage 2: Plan.
Ask for an outline, approach, criteria, or structure before drafting.
Stage 3: Produce.
Create the first usable version using the approved direction.
Stage 4: Critique.
Ask the AI to find weaknesses, omissions, unsupported claims, or likely objections.
Stage 5: Revise.
Make targeted changes and then perform human review.
For modern reasoning models, you do not always need elaborate instructions about internal reasoning. Focus on the goal, relevant context, constraints, and desired result.
Better Prompt Examples for Real Work
Using the notes below, draft a follow-up email to a potential customer who attended yesterday’s demo. Thank them, answer the two questions in my notes, and suggest a 20-minute follow-up call next week. Keep it under 180 words and professional but conversational. Do not invent pricing or product capabilities.
Meeting notes
Turn this transcript into a meeting brief with five sections: decisions made, action items with owners, deadlines, unresolved questions, and risks. Do not treat suggestions as decisions. If an owner or deadline was not stated, write “not assigned.”
Research
Research the current options for [topic]. Prioritize primary and authoritative sources. Compare the options by cost, major capabilities, limitations, and best-fit user. Cite each time-sensitive claim. Clearly separate facts from your recommendation.
Spreadsheet analysis
Analyze the attached sales spreadsheet. First check the columns and tell me whether anything appears missing or inconsistent. Then identify the five most important trends, unusual values, and changes from the prior period. Show the calculations behind any percentage you report.
Critique
Act as a skeptical customer reviewing this proposal. Identify the five places where the argument is weakest, unclear, unsupported, or likely to create an objection. Do not rewrite it yet. Rank the issues by how likely they are to affect the buying decision.
Powerful Prompting Techniques People Often Miss
Ask the AI to help write the prompt.
Describe your goal and ask, “What information would materially improve this request? Ask me up to five questions, then turn my answers into a reusable prompt.” OpenAI’s 2026 small-business resource provides a similar pattern for improving rough prompts. OpenAI Academy
Ask for options before choosing a direction.
Instead of accepting the first approach, request three meaningfully different options and the tradeoffs of each.
Ask it to identify what is missing.
Before generating an important deliverable, ask which missing facts, files, constraints, or decisions would most improve the result.
Ask for a rubric first.
For a proposal, article, plan, or strategy, ask what criteria should define a strong result. Review those criteria, then have the AI use them to evaluate the draft.
Separate creation from criticism.
Generate the draft, then switch jobs: “Now evaluate this as a skeptical reviewer. Do not defend the draft.” This can expose weaknesses that a simple rewrite request misses.
Make reusable prompts for repeated work.
If you write the same weekly report or customer response repeatedly, turn the successful instructions into a template with placeholders for the changing information.
Common Prompting Mistakes
Being vague.
“Help with marketing” does not define the actual job.
Expecting the AI to know private context you never supplied.
Tell it the relevant background or provide the approved source material.
Adding enormous amounts of irrelevant detail.
More context helps only when the context changes the answer.
Giving conflicting instructions.
“Be extremely detailed, but keep everything under 100 words” creates an unnecessary conflict.
Using only negative instructions.
Do not merely say what to avoid. Explain what the model should do instead. OpenAI’s prompt guidance recommends this approach. OpenAI Help
Starting over instead of correcting.
Use follow-up instructions to preserve what worked and fix what did not.
A Great Prompt Does Not Guarantee a Correct Answer
Prompt quality can improve relevance and usefulness, but AI can still make mistakes. Verify consequential facts, citations, calculations, policies, dates, legal claims, medical information, financial information, safety guidance, and anything else where an error could matter.
Prompting Improves the Assignment. Verification Protects the Result.
When possible, ask the AI to ground its work in authoritative source material. Then inspect the underlying source rather than assuming a polished answer must be accurate.
Detailed Prompts: Pros and Cons
| Pros | Cons |
|---|---|
| Reduce ambiguity and unnecessary guessing. | Can become bloated if every minor preference is included. |
| Make outputs more consistent and easier to review. | Overly rigid instructions can limit useful alternatives. |
| Can define tone, audience, format, and boundaries before generation. | Conflicting requirements can reduce quality. |
| Reusable prompts can save time on repeated work. | Old templates can become stale when the task or AI model changes. |
| Source-grounded prompts can reduce unnecessary invention. | Bad or outdated source material can still produce bad output. |
| Iteration lets you improve results without restarting. | Too much back-and-forth can waste time when the original goal was simple. |
A 7 Day Better Prompt Challenge
Day 1: Add a clear goal.
Replace one vague request with a specific job.
Day 2: Add context.
Tell the AI who the work is for and why it matters.
Day 3: Define the output.
Specify the format, length, tone, and structure.
Day 4: Give it a real source.
Use approved notes, a document, or data rather than asking it to guess.
Day 5: Add a boundary.
Tell it what not to invent and how to handle missing information.
Day 6: Critique and iterate.
Improve an answer with targeted follow-up instructions instead of restarting.
Day 7: Save a reusable prompt.
Turn one repeated work task into a template you can use again.
Final Thoughts
Writing better AI prompts is not about memorizing magic phrases.
It is about giving a capable tool a better assignment. Explain the goal. Supply the context that changes the answer. Give it the source material it should trust. Describe what the finished result should look like. Set boundaries where guessing would create problems.
Then review the response and continue the conversation. Correct misunderstandings. Ask for alternatives. Challenge weak reasoning. Provide missing information. Verify what matters.
The best prompt is not necessarily the longest prompt. It is the prompt that gives the AI enough of the right information to understand what successful work looks like.
Disclaimer: This article provides general educational and informational guidance about prompting artificial intelligence systems. AI capabilities, prompting behavior, product features, models, interfaces, privacy controls, and recommended practices change over time and vary by provider and product. Better prompting can improve usefulness but does not guarantee accuracy. AI outputs may contain errors, omissions, fabricated information, outdated claims, or inappropriate recommendations. Verify consequential information and follow your organization’s policies before providing confidential, proprietary, regulated, personal, legal, financial, health, security, or other sensitive information to an AI service. Use appropriately qualified professionals when specialized professional judgment is required. Product names and trademarks belong to their respective owners. This independently created article is not affiliated with or endorsed by any AI provider.
Official references: OpenAI Prompting Best Practices · OpenAI Academy Prompting · OpenAI Prompt Engineering Guide
