How to Actually Use Claude: 20 Ways to Make It Work Like You

Most people use Claude like a fancy search bar. This guide shows you how to set it up, personalize it, and turn it into a real thinking partner.

Most people open Claude, type one rushed question, get one decent answer, and think they are using AI.

Cute. But no.

That is like buying a full kitchen and only using the microwave.

Claude can do a lot more than answer random questions. It can help you think, write, plan, analyze, learn, build, edit, organize, and make better decisions. But it only gets truly useful when you stop treating it like a search bar and start treating it like a personalized work partner.

The difference is setup.

Claude has features like Projects, project instructions, profile instructions, styles, artifacts, memory, and extended thinking. Projects let you create focused workspaces with their own chat histories and knowledge bases, and project instructions let you tailor how Claude responds inside that workspace.

That matters because the best AI results do not come from magic prompts. They come from context. Claude needs to know who you are, what you are building, how you think, what you hate, what you love, and what kind of output actually helps you.

So let’s set it up like we mean it.

Here are 20 ways to actually use Claude.

1. Create Projects for Different Parts of Your Life

Do not dump your whole life into one endless Claude chat. That is how context gets messy, answers get weird, and your “business strategy” conversation starts remembering your grocery list.

Create separate Projects for your major lanes. You might have one for your business, one for writing, one for personal finance, one for travel planning, one for learning, and one for content creation. Claude Projects are designed as self-contained workspaces where you can upload documents, add context, and keep focused chats together.

The key is this: a Project is not magic memory. Claude does not automatically know every detail from every chat unless that information is added to the Project Knowledge or included in instructions. Anthropic’s own docs say context is not shared across project chats unless it is added to the project knowledge base.

Try this setup:

Create a Project for this area of my life: [business, writing, personal finance, travel, learning, etc.]

Help me decide what should belong in this Project, what files or context I should upload, what instructions I should add, and what types of chats I should keep separate.

2. Build a “User Manual” for Yourself

Claude gets better when it understands you. Most people skip this and then complain that the output feels generic. That is like hiring an assistant and refusing to explain the job.

Write a short user manual for Claude. Tell it who you are, what you do, what you are trying to build, how you like information, and what kind of responses make you want to throw your laptop into a lake.

Copy this prompt:

Help me create a personal user manual for Claude.

Ask me questions first about who I am, what I do, my goals, my work style, my communication preferences, my audience, my projects, and what I do not want in AI responses.

After I answer, turn my responses into a clear user manual I can add to my Claude Project Knowledge.

This is one of the highest-return AI moves you can make. Set it up once. Benefit every time.

3. Turn Your User Manual Into Project Instructions

Your user manual gives Claude context. Project instructions tell Claude how to behave.

That is the difference.

Project instructions should define the role Claude plays, the tone it should use, the formats you prefer, the things it should avoid, and the standards it should follow. Claude supports project instructions that apply across chats inside a Project.

Copy this prompt:

Based on my user manual below, write Project Instructions for Claude.

The instructions should tell Claude:
- Who I am
- What this Project is for
- How to communicate with me
- What tone and format to use
- What to avoid
- What quality standards to follow
- What questions to ask before complex tasks
- How to make outputs practical and specific

Write it as instructions Claude can follow. Keep it under 500 words.

Here is my user manual: [paste user manual]

Paste the final version into your Project Instructions. Now Claude has a default operating mode instead of starting every chat like a confused intern on day one.

4. Use Profile Instructions for Your Global Preferences

Project instructions are for one workspace. Profile instructions are broader.

Claude’s personalization features include profile instructions, project instructions, and styles. Profile instructions apply across your conversations, while project instructions are best for specific workflows or focused tasks.

Use profile instructions for things that are always true. For example, you might tell Claude to be concise, avoid corporate language, use plain English, ask clarifying questions before big tasks, and never bury the answer.

Copy this prompt:

Help me write account-wide profile instructions for Claude.

These should apply to all my conversations, not just one project.

I want Claude to know:
[describe your general preferences]

Write instructions that cover tone, structure, level of detail, what to avoid, and how to handle uncertainty.

This is how you stop correcting the same annoying habits over and over again.

5. Create a Custom Style From Your Own Writing

Claude can write. But without your style, it often writes like a polite robot that just discovered LinkedIn.

Fix that.

Claude lets users create custom styles from writing examples or by describing the style they want. The official docs say writing samples help Claude understand and match your preferred communication approach.

Give Claude three to five examples of your work. Then ask it to analyze the patterns. Do not just say, “Write like me.” That is too vague. Make Claude study your rhythm, sentence length, humor, vocabulary, paragraph style, and the stuff you never do.

Copy this prompt:

I want you to study my writing style so future drafts sound more like me.

Here are 3 to 5 writing samples:

Sample 1:
[paste]

Sample 2:
[paste]

Sample 3:
[paste]

Analyze my style in detail. Look at sentence length, rhythm, tone, vocabulary, humor, structure, paragraph length, openings, endings, emotional energy, and what I avoid.

Then create a reusable writing style guide I can save in Claude.

This is how you keep AI from sanding off your personality. Your weirdness is the asset.

6. Feed Claude Examples of Great Outputs

Writing samples show Claude your voice. Great output examples show Claude your standards.

This is different.

If you want Claude to help with sales pages, upload sales pages you love. If you want it to write newsletters, show it newsletters that work. If you want it to create reports, show it the kind of report you would actually send.

Copy this prompt:

I am going to give you examples of outputs I like.

Your job is to analyze what makes them effective. Study the structure, pacing, format, tone, level of detail, and reader experience.

After that, create a checklist you will use whenever you help me make something similar.

Examples:
[paste examples]

Claude performs better when it can see the target. Do not make it guess.

7. Stop Using Claude Like Google

Claude is not just a place to ask, “What is this?”

That is the lowest-value use.

Ask Claude to help you think through something. Give it context, goals, constraints, and a decision to help you make. The better question is not, “What is an email funnel?” The better question is, “I sell a $99 digital product to new creators. Help me design a five-email funnel that builds trust before asking them to buy.”

See the difference? One asks for a definition. The other asks for work.

Use this format:

I am trying to accomplish: [goal]

Here is the context: [details]

Here are my constraints: [time, budget, audience, tools, experience]

Here is what I need from you: [analysis, plan, draft, decision, critique, etc.]

Ask me any missing questions first. Then help me solve it.

Context is the difference between “fine” and “holy shit, this is useful.”

8. Make Claude Interview You Before It Starts

Most bad AI output comes from missing context. Claude fills the gaps with assumptions, then you waste time fixing them.

Make it ask questions first.

This works for writing, strategy, planning, research, product creation, sales copy, travel, budgeting, and basically any task with nuance.

Copy this prompt:

Before you start, ask me the 7 most important questions you need answered to do this well.

Do not create the final output yet.

After I answer, summarize what you learned, identify any remaining assumptions, and then begin.

This one habit will instantly improve your results. It makes Claude gather the ingredients before trying to bake the cake.

9. Ask Claude to Write the Prompt for You

You do not always need to know the perfect prompt. Claude can help you write it.

This is one of the easiest power moves. Tell Claude what you want done, then ask it to design the best prompt for that outcome.

Copy this prompt:

I need Claude to help me with this task: [describe task]

Write the best possible prompt for this task.

Include:
- The role Claude should play
- The context it needs
- The exact output format
- Important constraints
- Questions it should ask first
- Quality standards for the final result

Then ask me for any missing details before running the prompt.

Prompting is a skill. Let Claude teach you the skill while doing the work.

10. Use Claude as a Sparring Partner, Not a Cheerleader

Claude is very good at helping. Sometimes too good.

If you ask it to build on your idea, it will usually build. If you ask it to improve your plan, it will usually improve. But sometimes you need the opposite. You need Claude to challenge the idea before you spend money, time, or emotional energy on it.

Use Claude to stress-test your thinking.

Copy this prompt:

Here is my idea or plan: [paste plan]

Act as a ruthless strategic sparring partner.

Find the weak points, false assumptions, missing details, unrealistic expectations, and ways this could fail. Be direct and specific.

Then do the opposite. Build the strongest possible case for why this could work.

After both sides, give me your honest recommendation and the 3 changes that would make this plan stronger.

You do not need AI to clap for you. You need AI to help you win.

11. Use Extended Thinking for Hard Problems

Some tasks need quick output. Others need deeper reasoning.

Claude’s extended thinking is designed for complex tasks that benefit from more step-by-step reasoning, including math, coding, and analysis. (Claude API Docs)

Use it when the stakes are higher. Strategy, financial planning, business decisions, code architecture, legal research summaries, data analysis, and complex tradeoffs all deserve more thought.

Copy this prompt:

Think carefully before answering.

I do not need a fast answer. I need a useful one.

Analyze the problem, identify assumptions, compare options, explain tradeoffs, flag uncertainty, and then give me a clear recommendation.

Problem:
[paste problem]

Do not ask Claude to reveal hidden chain-of-thought. Ask for assumptions, tradeoffs, uncertainty, and a clear conclusion. That gets you the value without turning the answer into a messy reasoning diary.

12. Ask for Assumptions and Uncertainty

AI can sound confident even when it is guessing. That is dangerous and deeply annoying.

Build uncertainty into your prompts. Make Claude tell you what it knows, what it assumes, and what needs verification.

Copy this prompt:

Before giving me the answer, separate your response into:

1. What you know from the information I provided
2. What you are assuming
3. What you are uncertain about
4. What I should verify
5. Your best recommendation based on the current information

This is especially useful for money, health, legal, travel, business, and technical topics. Confidence is cute. Accuracy is better.

13. Control the Output Before Claude Writes Too Much

Claude will often give you more than you need. That sounds helpful until you are reading 900 words for a 30-word answer.

Tell it exactly what you want.

Length. Format. Tone. Number of options. No filler. No summary. No extra explanations.

Copy this prompt:

Answer in this format:

Length: [under 150 words / 5 bullets / 3 paragraphs / one-page plan]
Tone: [direct, warm, strategic, casual, etc.]
Format: [table, checklist, email draft, outline, decision memo, etc.]
Do not include: [preamble, summary, disclaimers, repeated context, generic advice]

Task:
[paste task]

This saves time, tokens, and your attention span.

14. Ban Preambles and Generic Filler

You do not need “Great question.” You do not need “Let me break this down.” You do not need a four-line summary of the question you just wrote.

Put that rule in your instructions.

Add this to Claude:

Do not start responses with affirmations, preambles, or restatements of my question.

Go directly to the answer.

Do not add a closing summary unless I ask for one.

Avoid generic advice. Give me specific, useful output.

This one is small but mighty. Claude gets sharper when you remove the fluff.

15. Start a New Chat When the Topic Changes

Long chats can get messy. The more you discuss, the more old context can influence new answers.

If you are switching from sales copy to travel planning, start a new chat. If you are switching from newsletter ideas to personal finance, start a new chat. Keep it inside the same Project if the broad context still matters, but stop dragging irrelevant history into unrelated work.

Use this rule:

New task, same project = new chat.

New life area = new project.

That little rule keeps Claude cleaner, faster, and less confused.

16. Use Project Knowledge Like a Brain Dump With Standards

Project Knowledge is where Claude gets truly personal. You can upload documents, text, code, and other files for Claude to use as context inside that Project.

For a writing Project, upload your voice guide, past essays, headlines, brand rules, audience notes, and content pillars. For a business Project, upload offers, FAQs, customer research, product details, SOPs, and launch plans. For a learning Project, upload notes, syllabi, papers, or study goals.

Copy this prompt:

I want to build the best possible Project Knowledge base for this Claude Project.

The Project is for: [describe purpose]

Tell me:
1. What documents I should upload
2. What context I should write from scratch
3. What examples would improve output quality
4. What instructions should live in Project Instructions instead
5. What information I should avoid uploading for privacy

Claude does not need your private passwords, account numbers, or sensitive identity details. It needs useful context. Keep it smart and safe.

17. Use Artifacts for Things You Want to Keep

If Claude creates something substantial, ask for it as an Artifact.

Artifacts are designed for standalone content like documents, code snippets, websites, SVGs, diagrams, flowcharts, and interactive components. They appear in a dedicated space separate from the conversation, which makes them easier to edit, reuse, and build on later.

Use Artifacts for things like content calendars, sales pages, dashboards, calculators, landing pages, course outlines, operating manuals, and decision frameworks.

Copy this prompt:

Create this as an Artifact so I can edit and reuse it later.

I want: [document, checklist, landing page, calculator, content calendar, workflow, etc.]

Purpose: [explain purpose]
Audience: [describe audience]
Style: [describe style]
Must include: [list requirements]

Chats are for conversation. Artifacts are for assets.

18. Learn Anything With the “Explain, Test, Adjust” Method

Claude is an incredible tutor if you stop asking for textbook explanations.

Ask it to explain a concept simply, test your understanding, then adjust based on your answer. This turns learning into a conversation instead of a lecture.

Copy this prompt:

Teach me [topic] like I am smart but new to it.

Use plain English, analogies, and everyday examples.

After each section, ask me one question to test whether I understood.

If my answer is weak, explain it a different way. If my answer is strong, go one level deeper.

Keep going until I can explain the topic back without jargon.

This works for AI, investing, coding, marketing, history, science, taxes, business models, and anything else that usually gets explained by people trying to sound important.

19. Use Claude as a Clarity Coach for Real Life

Claude can help you think through messy life stuff too.

Not as therapy. Not as a replacement for support, care, or real humans. But as a structured reflection tool when your brain feels like a browser with 47 tabs open.

Use it for decisions, conflict, priorities, overwhelm, habits, work stress, creative blocks, or anything you need to untangle.

Copy this prompt:

I want to think through something.

Do not give advice right away.

First, ask me questions to understand what is happening, what I feel, what I have already tried, what I am afraid of, and what outcome I want.

After I answer, reflect back what you hear. Then identify the real decision or tension underneath the situation.

Only then give me your honest perspective and 3 next steps.

Sometimes the win is not getting an answer. It is finally understanding the real question.

20. Stress-Test a Business Idea Before You Build It

This is one of my favorite ways to use Claude.

Most people fall in love with an idea too early. Then they spend weeks building something nobody asked for. Claude can help you pressure-test the idea before it eats your time, money, and confidence.

Copy this prompt:

I have a business idea I want to stress-test before I build it.

Here is the idea:
[paste idea]

Here is the target customer:
[describe customer]

Here is how it would make money:
[explain model]

Here is why I think it could work:
[explain reasoning]

Analyze this like a sharp business strategist.

Tell me:
1. The assumptions that could be wrong
2. Why the customer might not pay
3. Who already competes with this
4. What would make the idea hard to execute
5. What proof I need before building
6. The simplest test I could run in 7 days
7. What the idea would need to become to actually work

This is how AI saves you from expensive delusion. Not by killing your ideas, but by making them stronger before you bet on them.

Bonus: Build Custom Skills for Repeat Workflows

Once you know your repeat workflows, you can go deeper with Claude Skills.

Skills are folders of instructions, scripts, and resources that Claude can load for specialized tasks. Anthropic describes them as a way to improve consistency and performance for repeatable workflows, including brand guidelines, document creation, data analysis, and personal workflows.

This is more advanced, but it is powerful. A creator could build a Skill for newsletter editing. A business owner could build one for customer support replies. A teacher could build one for lesson planning. A media founder could build one for SEO article production.

Copy this prompt:

Help me design a custom Claude Skill for a repeat workflow.

The workflow is: [describe workflow]
I use it for: [describe purpose]
The output should always include: [list requirements]
The output should never include: [list avoidances]
My style or standards are: [describe standards]
Examples of good outputs are: [paste examples]

Create the Skill instructions, required resources, example inputs, example outputs, and rules for when Claude should use this Skill.

The goal is simple: stop repeating instructions that Claude could learn once.

The Point of These Prompts

Claude is not smarter than you. It does not replace your taste, instincts, lived experience, or ambition.

But it does have patience. It can hold context. It can compare options. It can critique ideas without getting tired. It can turn messy thoughts into clean drafts. It can help you learn faster, write better, plan smarter, and make decisions with more clarity.

That is the opportunity.

Most people will keep using Claude like a slightly better Google. They will ask one-line questions, accept average answers, and wonder why AI feels overhyped.

Do not do that.

Set up Projects. Add context. Create instructions. Build your style. Ask better questions. Use Artifacts. Stress-test your ideas. Make Claude interview you before it starts. Give it examples. Teach it your standards.

This is how Claude stops feeling generic.

This is how it starts working like you.

Start Here

Do not do all 20 steps at once.

Start with three moves: create one Project, write your user manual, and turn that into Project Instructions. Then add your writing samples and one workflow you want Claude to help with this week.

That is enough.

One smart setup can change the way you work every day.

Subscribe to Academy of AI for practical prompts, workflows, and strategies that help you use AI to work smarter, create faster, and build more freedom in your real life.