80+ AI Prompts and ChatGPT Prompt Commands: The Almost Complete Collection

By15ml Updated on
AI prompt encyclopedia thumbnail with bold black text and yellow design, introducing a collection of ChatGPT and LLM prompts.

Here is a collection of prompt commands I’ve gathered to help you get more out of AI. I organized them from the basics to more specific use cases, so bookmark this page and pull from it whenever you need it.

1. Basic Requests and Text Transformation

These are the prompts you’ll use most often. Use them to change the style of a text, summarize it, or rewrite it.

ELI11:

Ask the AI to explain a topic so simply that even an 11-year-old can understand it.

TL;DR:

(Too Long; Didn’t Read) Summarize a long article or document into a short version.

BRIEFLY:

Directly ask the AI to summarize only the key points in a concise way.

Humanize:

Make AI-like wording sound more natural and human.

Jargonize:

Rewrite ordinary sentences with more technical or professional terminology.

REWRITE AS:

Rewrite the given text in a specific style or format, such as an email or blog post.

2. Setting the Writing Format

Use these when you want to specify the AI’s direction, tone, and output format in more detail.

ACT AS:

Make the AI answer from the role of a specific job, person, or character.

AUDIENCE:

Specify the target audience for the answer, such as elementary school students, developers, or a fictional character.

TONE:

Specify the mood or tone of the writing.

FORMAT AS:

Ask the AI to format the result in a specific structure, such as a table, list, or JSON.

EXEC SUMMARY:

Write the answer as a business-style executive summary.

CHECKLIST:

Generate a checklist.

STEP-BY-STEP:

Generate a step-by-step guide.

BEGIN WITH / END WITH:

Specify how the response should begin and end.

GUARDRAIL:

Set rules or restrictions that must be followed.

PRIORITIZE:

Set priorities.

CONSTRAINT MODE:

Generate an output under specific constraints, such as character count limits or banned words.

FAILSAFE:

Prevent the answer from drifting too far from the topic or going in the wrong direction.

REQUIRES SOURCE:

Ask the AI to include sources or supporting evidence.

3. Deepening Thought and Analysis

Use these when a problem requires multi-angle, in-depth analysis.

LAYERED INSIGHT:

Analyze a topic in multiple layers, such as surface-level, structural, and deeper-level analysis.

DIALECTIC:

Explore a topic using the dialectical structure of thesis, antithesis, and synthesis.

CHAIN OF THOUGHT:

Show the reasoning process step by step until reaching a conclusion.

FIRST PRINCIPLES:

Break a problem down to its most fundamental elements and analyze it from the basics.

MULTI-PERSPECTIVE:

Analyze one issue from multiple stakeholder or perspective-based viewpoints.

SWOT:

Analyze strengths, weaknesses, opportunities, and threats.

COMPARE:

Compare and contrast two or more items based on specific criteria.

PRE-MORTEM:

Assume a plan has failed, then trace backward to identify and analyze the causes.

PITFALLS:

Find possible problems or logical traps in a specific plan or argument.

METRICS MODE:

Express or analyze results using measurable metrics or numbers.

Feynman Technique:

Explain a concept as if teaching it to someone else in order to improve understanding.

3-PASS ANALYSIS:

Analyze something in three passes: skim, identify the core, and review the details.

DO NOT SIMPLIFY:

Explain complex content without oversimplifying it, while preserving its complexity.

4. Special Work Modes

Switch the AI into a specific way of thinking or working mode.

SOCRATIC MODE:

Instead of giving the answer directly, keep asking questions so the user can find the answer on their own.

DEV MODE / PM MODE:

Answer using the perspective and terminology of a specific role, such as a developer or product manager.

THINK IN CODE:

Express the problem-solving process as actual code or pseudocode.

DRAFT | REVIEW | PUBLISH:

Imitate a writing workflow and improve the output step by step.

REFLECTIVE MODE:

Review, analyze, or evaluate a previous conversation or the AI’s own answer.

BLACK_BOX_:

Infer or analyze a system using only its inputs and outputs, without knowing its internal process.

5. AI Self-Verification and Revision

Use these to improve the reliability of AI answers and check for bias.

EVAL-SELF:

Have the AI evaluate how accurate and useful its own answer is.

SYSTEMATIC BIAS CHECK:

Have the AI review potential biases that may be included in its answer.

CRITIC_LOOP:

Have the AI act as its own critic and repeatedly review and improve its output.

TRUTH_GATE:

Ask the AI to go through a process of verifying whether the generated information is factual.

SELF_PATCH:

Have the AI correct and supplement its own answer if it finds errors.

6. Context and Memory Management

Use these to manage or reset the context of a conversation.

CONTEXT STACK:

Ask the AI to summarize the conversation context so far.

USE PRIOR / RECALL:

Bring back and use a specific part of a previous conversation.

ZERO SHOT:

Perform a task immediately without examples or prior information.

ZERO-KNOWLEDGE PROMPTING:

Reason only from the given information, without relying on prior knowledge or assumptions.

NO DEFAULTS:

Push the AI to minimize its default behavior, such as excessive politeness, and give a more raw answer.

THOUGHT_WIPE / ZERO_IMPRINT / NO_TRACE:

Forget the previous conversation context and start from a completely fresh state.

7. Simulation and Interaction

Create fictional scenarios and encourage interaction.

FRICTION SIMULATION:

Simulate a situation involving friction or conflict.

IF/THEN INTERACTIVE:

Create a scenario where different outcomes appear depending on the user’s choices.

PRISONER_DILEMMA:

Simulate a prisoner’s dilemma situation.

MULTI-AGENT SIMULATION:

Simulate a situation where multiple fictional agents or roles interact with one another.

8. Fun & Miscellaneous Uses

These are prompt commands for fun.

AITAH (Am I the Asshole?): Describe a specific situation and ask whether you were in the wrong.

HOTORNOT?: Predict or evaluate how people might react to an idea, design, or similar item.

9. Advanced & Experimental Prompts

EN draft → KR adapt → KR-QA:

EN draft → KR translation → KR QA check:

FORCE TRACE:

Trace backward to identify the factors, such as data or rules, that had the greatest impact on a specific result.

FORCE TRACE:

Trace backward to identify the factors, such as data or rules, that had the greatest impact on a specific result.

LATENT ROLE SWITCHING:

Guide the AI to subtly switch roles during a conversation without explicit instructions.

ECHO-FREEZE:

Freeze an answer or concept from a specific point and dig deeply into only that target.

REASONING ANCHORS:

Use a specific fact or rule as an anchor and force all reasoning to be based on it.

INSTRUCTION COLLAPSING:

Collapse multiple complex instructions into one condensed command and execute it.

FORK_CONTEXT:

Branch the current conversation into several possible scenarios and explore each one.

QUANT_CHAIN:

Break thinking into extremely small units and connect them to perform complex reasoning.

10. Other Categories (More to Be Added)

There are many other prompts for specific purposes as well. New prompts will continue to be added over time.

ECHO_LOOP:

Repeatedly analyze, summarize, and reinterpret.

NULL_INJECTION:

Guide the AI to think from a perspective where a specific concept or element has been removed.

MODEL_REFLE:

Have the model check its own answer.

SHADOW_PRO:

Use when you want the AI to implicitly profile the user.

REFLECTION PRIMING:

Prime the model to give deeper and more reflective answers.

EXTERNAL SIMULATION BLOCK:

Run a simulation based on an assumed external system.

TEMPERATURE_SIM:

Specify the model’s creativity level metaphorically.


PARALLEL LENSES (Use when you want one topic analyzed in parallel from multiple perspectives.
NO AUTOPILOT (Make the AI think freely without following the rules established so far.)
DELIBERATE THINKING (Prompt the AI to think carefully before making a judgment.)
SCHEMA (Convert into a schema structure.)
ROLE: TASK: FORMAT: (role, task, format)
TOKEN_OVERF (Generate a closing line or final flourish.)
CROSS-MODEL BLEND (Blend the styles of different models.)
TOKEN MASKING (Mask specific words]
PSEUDO-SCRIPTING
TRIGGER_CHAIN
SAFE_LATCH (safe mode)
RAW_COMPILE (Output the draft as-is in a raw state.)
AUTO_MODULATE (Ask the AI to automatically adjust tone, detail level, and response style based on context.)

추천 포스트