How to Use Claude AI for Writing, Long Documents and Projects
Short answer: To use Claude AI, sign up at claude.ai, start a chat, and give it a clear task with your files attached. For work you repeat, create a Project: add standing instructions and reference files once, and every chat inside it starts with that context. Claude is especially good at long documents and careful, natural writing.
Claude is an AI assistant made by Anthropic. It does many of the same things as ChatGPT, and the chat box looks familiar, but people who switch usually notice two differences quickly: it handles long documents patiently, and its writing sounds less like a template.
This tutorial covers the first-day setup, then the feature that pays off most for working people: a Project for a recurring task. We build one for a weekly client report, with the full instructions text you can copy and adapt, and finish with the long-PDF habits we rely on in agency work.
What is Claude AI, and how is it different from ChatGPT?
Claude is a general AI assistant for writing, analysis, coding and reading files. Compared with ChatGPT, it leans towards long-document work and writing quality, while ChatGPT offers a wider set of built-in extras such as image generation.
| Task | Claude | ChatGPT |
|---|---|---|
| Reading a long PDF or several files at once | A core strength; large context and careful quoting | Capable; better on paid plans |
| Writing in a natural, specific tone | Often needs fewer rewrites | Good, but defaults can sound generic |
| Creating images | Not a built-in feature as of November 2026 | Built in |
| Building small interactive tools in the chat | Artifacts | Canvas and code tools |
| Recurring work with saved context | Projects | Projects |
Neither is better for everything. For a fuller task-by-task view, see our ChatGPT, Claude and Gemini comparison.
How do you start using Claude? (first 10 minutes)
- Go to claude.ai or install the Claude app, and sign up with email or Google.
- Stay on the free plan to begin. Usage resets every few hours, and long files use up your allowance faster. Paid plans raise the limits: as of November 2026, Anthropic bills Indian users in rupees, with Pro at about ₹2,000 a month and Max from about ₹11,999 a month (GST included). Check the current price on Anthropic's site before you pay.
- In Settings, add a short profile of who you are and how you like answers (for example: 'I run social media for small Delhi businesses; give me short, practical answers in Indian English').
- Start a chat with a real task, not a test question. Attach the file the task depends on using the paperclip or by dragging it in.
- Read the answer and reply with one precise correction at a time. 'Shorter, and drop the second paragraph' works better than 'make it better'.
- Start a fresh chat when the topic changes. Long mixed chats use more of your limit and blur the context.
If your first prompts feel hit-and-miss, the problem is almost always missing context. The checklist in how to write AI prompts works in Claude unchanged.
What are Claude Projects, and when should you use one?
A Project is a workspace with its own instructions and files. Every chat you start inside it already knows those instructions and can read those files, so you stop re-explaining the same background each week.
Use a Project for anything you do more than twice with the same context: a weekly report, a client's brand voice, a course you are studying, a set of HR policies. As of November 2026 Projects are available on Claude's plans with limits on the free tier; check what your plan includes.
- Click Projects in the sidebar, then create a new Project. Name it after the job, such as 'Weekly report: Dwarka dental clinic'.
- Open the Project's instructions field and paste your standing instructions (an example follows below).
- Add reference files to the Project knowledge: last month's report you liked, the client's service list, a glossary of terms.
- Each week, start a new chat inside the Project and attach only that week's data.
- When you correct the same thing twice, move that correction into the instructions.
Worked example: a Project for a weekly client report
Say you handle Instagram and Google Business Profile for a dental clinic in Dwarka, and every Monday you send the owner a short report. Here is the instructions text for that Project, written the way we would brief a junior team member:
'You help me write the weekly marketing report for a dental clinic in Dwarka, New Delhi. The reader is the clinic owner: a busy dentist, not a marketer. Each week I will attach a CSV export of Instagram insights and a screenshot or CSV of Google Business Profile performance. Write the report in this order: 1) Three-line summary of the week in plain English. 2) A table: metric, this week, last week, change. Use only these metrics: reach, profile visits, website clicks, calls from Google, direction requests. 3) What worked: one or two posts, with the reason in one line each. 4) What to try next week: two specific actions. Rules: use only numbers present in the attached files; if a number is missing, write Not available, never estimate. Calculate change as a percentage rounded to one decimal place and show your working in a short note at the end. No jargon such as engagement rate or impressions without a one-line explanation. Keep the whole report under 250 words plus the table. Indian English, rupees with the ₹ sign.'
Why each line is there: naming the reader stops marketing jargon; fixing the metric list stops the report changing shape every week; 'never estimate' prevents invented numbers, which is the most dangerous failure in client reporting; asking for the working lets you check the maths in ten seconds instead of trusting it.
One practitioner habit: check two or three numbers against the source export every single week, even after months of clean reports. AI arithmetic on CSVs is usually right and occasionally confidently wrong, and the client will only remember the wrong one.
How do you get reliable answers from long PDFs in Claude?
Claude can read long files, but reliable answers still depend on how you ask. These are the habits that hold up in agency use with tenders, policy documents and brand guidelines.
- Ask for page references with every claim: 'Cite the page number for each point.' Then spot-check two of them.
- Ask questions section by section for anything over about 50 pages, rather than 'summarise this whole document'.
- Ask it to quote exact wording for clauses that matter (penalties, eligibility, deadlines), not paraphrase them.
- Tables in PDFs are the weakest spot. For a pricing table, ask Claude to reproduce the table first, compare it with the PDF, then ask questions.
- Scanned PDFs (photos of paper) read less reliably than digital ones. If text cannot be selected in your PDF viewer, expect more errors.
- Ask 'What does this document not say about X?' Absence is where summaries most often mislead.
- Remove personal or client-confidential details you do not need before uploading, whatever the plan.
What do beginners get wrong with Claude?
- Treating it as a search engine. Claude can search the web when that tool is on, but it is strongest when you give it the material to work from.
- Writing one giant prompt with five tasks. Split them; quality drops when one message asks for a summary, an email and a table together.
- Never using Projects, so every Monday starts from zero.
- Ignoring Artifacts. When you ask for a calculator, checklist or small web page, Claude can build it in a side panel; our Claude Artifacts tutorial walks through two builds step by step.
- Accepting numbers without checking. Claude writes confidently even when it is wrong, like every AI assistant.
Claude, ChatGPT, Gemini and NotebookLM are all covered hands-on in our Applied AI Professional course, with each tool used on the kind of weekly work shown above.
Frequently asked questions
Is Claude AI free to use?
Yes, Claude has a free plan with usage limits that reset every few hours. Paid plans (Pro, and Max for heavy use) raise the limits and add features; as of November 2026 Pro costs about ₹2,000 a month in India, but check Anthropic's site for current prices.
What are Claude Projects?
Projects are workspaces with their own instructions and reference files. Every chat inside a Project starts with that context, which suits recurring tasks like weekly reports or a client's brand voice.
Can Claude read PDFs?
Yes. You can upload PDFs and other files and ask questions about them. For long files, ask for page references and check a few, because summaries can still miss or misread details.
Is Claude better than ChatGPT?
It depends on the task. Claude is often preferred for long documents and natural writing, while ChatGPT offers more built-in extras such as image generation. Many professionals use both.
Can Claude generate images?
As of November 2026, Claude does not generate images as a built-in feature. It can describe images, read screenshots and build diagrams or simple graphics as code inside Artifacts.
Keep reading in AI tools
- How to Write AI Prompts That Work First Time Start here
- Perplexity vs ChatGPT for research: which one to trust for what
- How to Use NotebookLM to Study and Research From Your Own Files
- Is Prompt Engineering a Good Career in India?
- Best Free AI Image Generators in 2026, and How to Test Them on Indian Images
Related: Applied AI Professional · Claude Artifacts Tutorial: Build Two Working Tools Without Coding · All guides in AI tools