Prompts for ChatGPT
Build clear, reusable prompts for ChatGPT, then start from 100 practical prompt examples for real tasks across work, writing, research, files, data, coding, images, and planning.
Shape the request
Your editable prompt
Review the generated structure, edit anything you want, then copy it or open ChatGPT in a separate tab. Your prompt is never auto-submitted.
Open ChatGPT only navigates to ChatGPT after your click. Mycelgrid does not automatically send the prompt or any source material.
Give ChatGPT the task, the context, and the shape of the answer
A strong prompt reduces ambiguity. Start with the task, add the background ChatGPT needs, say who the answer is for when that matters, define how sources or tools should be used, and specify the output you want.
Lead with the task
Use a direct verb such as summarize, compare, analyze, draft, explain, debug, plan, or critique. State what a successful result needs to do.
Add only useful context
Include purpose, audience, background, constraints, and source-use instructions that change the answer. More words are not automatically a better prompt.
Define the output
Ask for bullets, sections, a table, a draft, structured data, or a step-by-step plan. Add quality checks when accuracy or source fidelity matters.
100 prompt examples for real tasks
Search prompts, browse by task, or choose a Quick start. Every example loads a complete, editable configuration into the builder so you can customize it for your own work.
100 ChatGPT prompt examples for real tasks
A durable structure for better prompts
OpenAI’s current prompting guidance emphasizes clear, specific requests, enough context, right-sized tasks, and iterative refinement. Mycelgrid turns those principles into a reusable structure and adds explicit source-use, output, and quality-check instructions when the task needs them.
Task
Say what ChatGPT should do and what a successful result needs to accomplish.
Context and audience
Add the purpose, background, audience, and domain constraints that materially change the answer.
Sources and tools
Tell ChatGPT what material to use. When current information, files, images, code, or data matter, say what to inspect and what evidence boundary to respect.
Constraints
Specify required content, exclusions, tone, length, decision rules, or success criteria only when they affect the result.
Output and quality check
Define the response shape, then ask ChatGPT to flag uncertainty, conflicts, unsupported claims, missing inputs, or calculations that need checking.
Iterate
For complex work, use the first useful answer as a working draft. Refine the task, context, or evidence boundary with focused follow-up prompts.
Tell ChatGPT what to inspect, what to preserve, and what not to invent
When your task depends on evidence, define the source boundary. This builder can prepare instructions for pasted text, documents, PDFs, images, screenshots, spreadsheets, code, data, or current web information without uploading any of those materials to Mycelgrid.
Documents and text
Ask ChatGPT to distinguish what the source explicitly supports from interpretation. Request section, page, row, or passage references when the material makes them available.
Images and screenshots
Specify the visual details to inspect: objects, text, layout, composition, differences, or evidence relevant to the task. Ask it to separate observation from interpretation.
Data, code, and current web information
Name the files, functions, columns, calculations, or facts that matter. When web verification is needed and available, ask for current sources and citations rather than unsupported claims.
Specific beats vague
The useful change is not simply making a prompt longer. It is giving ChatGPT the information and constraints that materially define success.
Vague request
This leaves the purpose, audience, evidence boundary, and format unspecified.
Analyze this customer feedback.
Structured request
This version defines the task, source boundary, decision context, and output.
TASK Analyze the customer feedback I provide and identify the recurring themes that should affect our product priorities. CONTEXT We are preparing the next quarterly roadmap. Focus on problems users repeatedly mention, not one-off feature requests. SOURCE USE Use the provided feedback as the primary evidence. Do not invent frequencies that are not supported by the data. OUTPUT FORMAT Return: 1) top themes, 2) representative evidence, 3) severity and frequency signals, 4) suggested next questions. QUALITY CHECK Flag weak evidence, conflicting feedback, and any conclusion that needs validation.
A better prompt can improve the request; it cannot guarantee a correct answer
ChatGPT can still produce incomplete, inaccurate, or unsuitable output. Review important claims, calculations, citations, code, and high-stakes recommendations before acting on them.
Local prompt building
Mycelgrid constructs the prompt in your browser. This page does not upload your builder inputs or automatically send them to ChatGPT.
Verify important output
Use source-first or strict-source instructions when fidelity matters. For professional, legal, medical, financial, or safety-critical work, use qualified human review.
Capabilities vary and change
File, image, data-analysis, web, and other capabilities can vary by plan, model, workspace, and product updates. This page avoids hard-coding model numbers, context limits, or feature guarantees.
Questions about prompts for ChatGPT
What are the best prompts for ChatGPT?
The best prompt is the one that fits the task. A useful ChatGPT prompt clearly states the task, gives relevant context, defines the audience or role when needed, explains how source material should be used, and specifies the desired output. Important tasks also benefit from a quality-check instruction that asks ChatGPT to flag uncertainty or missing evidence.
Should ChatGPT prompts be long?
Not necessarily. A short prompt can work well when the task is simple and the context is obvious. Add detail when it changes the answer: purpose, constraints, evidence boundaries, audience, format, and success criteria.
Can I build prompts for documents, PDFs, images, code, or data?
Yes. The builder can prepare instructions for those source types. Mycelgrid does not upload the source itself; you attach or paste the material in ChatGPT when you choose to use the prompt. Availability of file, image, and data-analysis features can vary by plan and workspace.
Can I make a prompt that asks ChatGPT to search the web?
Yes. Choose current information from the web as the working material or use the verification preference. The generated prompt asks ChatGPT to use current sources when web access is available and to avoid pretending it verified something when it could not.
Does Mycelgrid send my prompt to ChatGPT?
No. The prompt is built locally in your browser. Copy Prompt uses your clipboard only after you click it. Open ChatGPT opens ChatGPT in a new tab but does not automatically submit the prompt.
What should I do if ChatGPT gives a weak answer?
Refine the task, add missing context, narrow the requested output, or ask a focused follow-up question. If accuracy matters, provide the relevant source material and tell ChatGPT how to handle uncertainty or unsupported claims.