Heads up: some links below are referral links. They cost you nothing extra and never change what we recommend. How this site makes money.
Most "how to use Claude" guides spend a thousand words explaining what artificial intelligence is. You already know. This one covers the things that actually change the quality of what you get back.
Getting started
Sign up for a free account, and that is genuinely it — there is nothing to install and no configuration. Type in the box.
The free tier gives you enough usage to decide whether Claude suits your work. Do not pay before you have hit its limits; if you never hit them, you have your answer.
Four habits that change your results
1. Say who it is for
"Explain compound interest" gets you an encyclopedia entry. "Explain compound interest to my 14-year-old, who finds maths boring" gets you something usable. The audience shapes register, length, vocabulary and examples all at once, and it costs you six words.
2. Give it your material
Claude's biggest practical advantage is how much it can read at once. Paste in the actual document, the actual error message, the actual draft. Describing your writing style gets you a guess; pasting three paragraphs of it gets you a match.
3. Ask for a plan first on anything big
For a substantial task, ask for an outline before the draft. You will spot the wrong direction in a ten-line outline in seconds, where you would have to read two thousand words to spot it in a draft. Approve the outline, then ask it to write.
4. Correct it directly
"Too formal, and the second section is wrong — we launched in March, not June" works better than starting over. Claude keeps the thread. Treating it as a conversation with an assistant who is doing a reasonable job produces better results than treating it as a vending machine that gave you the wrong crisps.
The one-line version: tell it who the reader is, give it your real material, agree a plan before a long draft, and correct it in place rather than restarting.
Working with documents
Upload a file and ask questions about it. This works on long documents — contracts, reports, transcripts, research papers — and it is where Claude is most obviously ahead.
Useful patterns:
- "Summarise this in five bullets for someone who has ten seconds." Forces genuine prioritisation instead of a shorter version of everything.
- "What is missing from this?" Better than "review this", which invites praise.
- "What would a sceptical reader push back on?" The single most useful question you can ask about your own draft.
- "Find every date and commitment in this contract and list them." Extraction is unglamorous and it is where the hours actually go.
Projects
If you keep re-explaining the same background — your company, your style, your codebase — Projects let you set that context once so it applies to every conversation inside them.
Worth setting up for anything you do repeatedly. A "client work" project holding your tone-of-voice notes saves you the same paragraph of preamble fifty times.
Common mistakes
Being vague and hoping
Claude does what you asked rather than guessing what you meant. That is a feature once you adjust, but it does mean woolly prompts get literal-minded answers.
Accepting the first draft
The first output is a starting point. The second, after you have said what is wrong with the first, is usually the one worth keeping.
Trusting facts without checking
Claude can state something false with complete confidence, particularly about recent events, specific numbers and anything with a date on it. For research that has to hold up, verify with a tool built for it — Perplexity cites its sources, which makes checking a single click.
Not asking it to disagree with you
Ask "what is wrong with this plan?" and you get a far more valuable answer than "what do you think of this plan?". The framing does most of the work.
Ready to try it
All of the above works on the free tier. Pick a real task rather than a test one — that is the only way to find out whether it saves you time.