Claude.ai Basics
Claude.ai fundamentals: models, plans, core features
Last reviewed
Recommended
Claude.ai Basics — Timed Test 4 (10 questions)
No account needed. Answers and explanations arrive when you submit.
What this topic tests
The mix every Claude.ai Basics set is built to, and the questions published against it so far. Nothing here is hidden before you start.
| Level | Target share | Published |
|---|---|---|
| Easy | 50% | 25 |
| Medium | 34% | 17 |
| Hard | 16% | 8 |
| Total | 50 |
Claude.ai Basics — the theory
What Claude Is
Claude is an AI assistant made by Anthropic. You type a question or a task in plain language, Claude answers, and you keep going from there — a conversation rather than a search box. Most people use it at claude.ai in a browser, and the same account works in the Claude desktop and mobile apps.
Claude is good at working with text and ideas: drafting and editing writing, summarising a long document, explaining something you do not understand, planning, brainstorming, translating, and working through code. It is not a search engine and it is not a database of facts. It is a model that has read a great deal and reasons about what you put in front of it, which is why giving it the right material matters more than phrasing your question perfectly.
Chats, and How a Conversation Works
Everything happens inside a chat. Each chat keeps its own history, so Claude can see what you said earlier in that conversation and refer back to it. Start a new chat when you change subject — a fresh chat gives Claude a clean slate, and a long chat that wanders makes its answers worse rather than better.
Your chats are saved in the sidebar, and you can search them to find something you discussed before. Two things are worth knowing early: a very long conversation eventually gets harder for Claude to keep track of, and each chat is separate by default. What you told Claude on Monday is not automatically available on Tuesday unless you turn on Memory, which is covered below.
Plans and Limits
Claude has a free tier and several paid plans — Pro and Max for individuals, Team and Enterprise for organisations. The free tier is a real product, not a trial: you can hold conversations, upload files and create projects on it. Paying mainly buys you more usage, plus extra surfaces such as Claude Code and Claude Cowork.
Usage is the part people misunderstand. Anthropic describes usage limits as controlling how much you can interact with Claude over a specific time period, rather than as a fixed number of messages. How fast you consume that allowance depends on the length and complexity of your conversations, which features you use, which model you are chatting with, and the effort level you have selected. A short question costs far less than a long document analysis.
One rule catches people out: usage across all the Claude surfaces — claude.ai, Claude Code and Claude Desktop — counts toward the same limit. Spending the morning in one and the afternoon in another does not give you two allowances.
File Uploads
You can attach files directly to a chat and ask Claude about them. Documents include PDF, DOCX, CSV, TXT, HTML, RTF, ODT, EPUB, JSON and XLSX; images include JPEG, PNG, GIF and WebP. In a chat you can attach up to 20 files, up to 500MB each, and a PDF can be up to 1,000 pages. Files added to a project's knowledge base have a smaller 30MB limit each.
This is the single fastest way to get a better answer. Claude cannot see a document you have not given it, and pasting or attaching the actual text beats describing it — including for anything too recent for the model to know about on its own.
Projects
A project is a workspace. Anthropic describes projects as self-contained workspaces with their own chat histories and knowledge bases, and the knowledge base is the point: you upload the documents, notes or code that a piece of ongoing work depends on once, and every chat inside that project can draw on them without you re-attaching anything.
Projects are available to everyone, including free accounts, where you can create up to five. Use one whenever you will come back to the same subject more than once — a course you are studying, a report you are drafting, a codebase you are learning.
Artifacts
When Claude produces something substantial and self-contained — a document, a piece of code, a small app — it opens in a separate panel beside the conversation instead of scrolling past inside it. That panel is an artifact. Anthropic's rule of thumb is content that is significant and self-contained, typically over fifteen lines, and something you are likely to want to edit, iterate on or reuse outside the chat.
The practical benefit is that you can keep refining the thing while the conversation continues around it, then copy, download or share the result.
Memory
By default Claude starts each conversation fresh. Memory changes that: with it switched on, Claude generates memory from your chats and can carry useful context forward, so you are not reintroducing yourself and your work every time.
It is yours to control. You turn it on under Settings > Memory, and from the same place you can pause it — keeping what exists but adding nothing new — or reset it, which permanently deletes what has been remembered. If you would rather Claude forgot a particular conversation, that is a decision you make, not one made for you.
Basic Prompting
You do not need special syntax. A few habits do most of the work:
- Say what you want, and who it is for. "Summarise this in five bullets for a client who has not read it" beats "summarise this".
- Give Claude the material. Attach the document rather than describing it.
- Ask for a format. A table, a checklist, three options, 200 words.
- Iterate rather than restart. "Shorter, and drop the third point" is faster than rewriting the whole request.
- Ask it to think it through on hard problems, and to show the reasoning you want to check.
Claude Prompting Best Practices and Prompt Engineering Fundamentals go deeper.
Safety and What to Check
Two habits protect you. First, verify anything that matters. Claude's knowledge stops at a cutoff date, and it can state something wrong with complete confidence. For anything consequential — legal, medical, financial, or a fact you are about to publish — check the source. If the subject is recent, give Claude the current material rather than trusting recall.
Second, be deliberate about what you paste. Treat other people's personal data, credentials and confidential documents the way you would in any other cloud tool.
There are limits on what Claude will help with, set out in Anthropic's Usage Policy. Beyond outright prohibited uses, the policy also attaches obligations to lawful but high-risk uses: if you are building something that gives advice or makes decisions directly affecting individuals, a qualified professional has to review the output before it goes out, and a consumer-facing chatbot has to disclose that people are talking to an AI.
Where to Go Next
Projects & Artifacts covers the two workspace features in depth. If you are weighing Claude against the obvious alternative, ChatGPT Basics covers the same ground for that product. Claude Code Basics is Claude in the terminal for developers, and AI Agents Basics covers what happens when an assistant is given tools and left to work.
Sources
All read 2026-08-09. What are projects? · What are artifacts? · Chat search and memory · Upload files to Claude · How usage and length limits work · Plans & pricing · Usage Policy
Sample questions
Three questions from this topic, with the answer and the reasoning shown.
Q1EasyAnthropic's getting-started guidance centres on one habit for writing a good request. What is it?
- Keep every request under a single short sentence
- Be specific — the more specific the request, the better Claude can helpCorrect
- Use particular keywords and phrases that the model has been specially tuned to recognise
- Always begin by telling Claude which model it is running as
Explanation
The principle — most disappointing answers are answers to a vaguer question than the one you meant.
Why the key is correct — Anthropic's guidance for new users states it directly: be specific, because the more specific your request, the better Claude can assist you. The companion advice is to start simple and refine from there.
Why the others are wrong — brevity is not the objective, there are no magic keywords, and naming the model in your message changes nothing.
Remember this — say what you want, who it is for, and what form it should take.
Sources — Anthropic's getting started guide.
Q2EasyBy default, what does Claude carry over from your earlier conversations into a brand-new chat?
- Your preferences only, with the subject matter left behind
- Everything, since all your chats form one continuous conversation
- The most recent chat only, which is always kept available
- Nothing, unless you have switched memory onCorrect
Explanation
The principle — separateness is the default, and continuity is something you opt into.
Why the key is correct — a new chat starts without the others. Anthropic describes memory as a capability Claude can now generate from your chats, which is precisely the feature that changes this behaviour once enabled.
Why the others are wrong — preferences are not exempt, the chats are not one thread, and the previous conversation is not silently attached.
Remember this — a fresh chat is a fresh start. If you want continuity, turn memory on or work inside a project.
Sources — Anthropic's chat search and memory article.
Q3EasyClaude answers, and part of the response opens in a panel beside the conversation. What is that panel?
- A preview of a file that has been downloaded to your computer
- An artifact — standalone content shown in its own windowCorrect
- An advertisement panel that paid plans remove
- A private scratchpad showing Claude's internal reasoning
Explanation
The principle — substantial output gets its own surface so it does not scroll away inside the conversation.
Why the key is correct — Anthropic describes Claude sharing substantial, standalone content in a dedicated window separate from the main conversation. That window is the artifact.
Why the others are wrong — nothing has been downloaded, nothing is being advertised, and the panel is not a view of the model's reasoning.
Remember this — if it appeared beside the chat, it is meant to be worked on rather than merely read.
Sources — Anthropic's artifacts documentation.
Practise all 50 questions
Every published question in Claude.ai Basics, with its answer and explanation.
- Anthropic's getting-started guidance centres on one habit for writing a good request. What is it?easy
- By default, what does Claude carry over from your earlier conversations into a brand-new chat?easy
- Claude answers, and part of the response opens in a panel beside the conversation. What is that panel?easy
- Claude has built a checklist in a side panel and you want it as a file on your own computer. What does that panel let you do?easy
- How does Anthropic describe the usage limits on a Claude plan?easy
- On the free plan, Claude tells you at eleven in the morning that you have reached your limit. What governs when you can carry on?easy
- Three weeks ago, on your Pro plan, you worked out a pricing approach with Claude and cannot remember which conversation it was in. What can you do?easy
- What can you do on Claude's free tier?easy
- What does Anthropic advise about typing highly sensitive personal details into a Claude conversation?easy
- What does switching web search on change about how Claude answers?easy
- What does upgrading from the free tier to a paid Claude plan mainly give you?easy
- What is a project in Claude?easy
- What is the point of a project's knowledge base?easy
- What kind of content does Claude place in an artifact rather than inline?easy
- What kind of tool is Claude?easy
- What limits apply when you attach files to a single Claude chat?easy
- Where do you control Claude's memory, and what can you do there?easy
- Which factors change how quickly you use up your Claude allowance?easy
- Which kinds of file can you attach directly to a Claude conversation?easy
- Why should you verify a Claude answer before relying on it for something consequential?easy
- You ask Claude about a phone released last month and it describes the wrong specifications, without hesitating. What is the most likely explanation?easy
- You have a free Claude account and want to organise ongoing work into workspaces. What is possible?easy
- You paste a long article into a chat and ask Claude for the key points. What is Claude actually doing?easy
- You spend an hour in an incognito chat drafting a speech, close it, and come back the next day for the draft. What has become of it?easy
- You work in the claude.ai web app in the morning and Claude Code in the afternoon. How is that usage counted?easy
- Claude has produced a two-page briefing note in a side panel. You want the same note, but shorter and with a different opening. What is the efficient next step?medium
- Claude states a statistic you intend to put in a published article. What is the sound way to verify it?medium
- Halfway through a very long analysis Claude says you have reached a limit, yet a brand-new chat you open straight afterwards works normally. What did you actually run into?medium
- Memory is on and Claude has built up a useful picture of how you work. For the next fortnight you are on an unrelated confidential matter and want nothing new recorded, while keeping everything already there. What do you choose?medium
- Two colleagues are on identical paid plans. One sends forty short questions a day and never hits a limit. The other sends six requests a day, each analysing a long report, and is cut off by mid-afternoon. What explains the difference?medium
- You are stuck on a logic puzzle with all the information already written down in front of you. Which capability suits the task?medium
- You are writing a dissertation over three months, returning most days, always working from the same twelve source papers and a style guide. How should you set this up in Claude?medium
- You ask Claude to "make this email better" and get back something polite but generic. What change to the request is most likely to help?medium
- You attach a 300-page PDF and ask Claude to interpret the charts in it. Claude discusses the surrounding wording confidently but says almost nothing about the figures themselves. Why?medium
- You need Claude's help with a 60-page contract that was emailed to you as a PDF. What gets the most reliable answer?medium
- You need to know what yesterday's inflation figure was. Which Claude capability fits the task?medium
- You told Claude about your writing style last week in a chat you have since closed. Today you open a new chat and want that style applied again, without turning memory on. What works?medium
- You use Claude twice a week for short questions and have never seen a usage limit. A colleague insists you should be paying. What is the sound reasoning?medium
- You want the current share price of a company, accurate to the minute, to place a trade. How should you approach this?medium
- You want to compare twenty-five scanned reports in one conversation. Each is a 40MB PDF of about 300 pages. What should you expect?medium
- You want to talk through something you would rather was not used to improve future Claude models, but you would prefer to leave your usual privacy settings alone. What achieves that?medium
- Your team's brand guidelines change roughly once a quarter, and you reference them in most conversations about copy. Where should they live?medium
- A clinic wants Claude to draft letters advising patients on treatment options. A nurse will read and sign each letter before it is sent. Under Anthropic's Usage Policy, how should this be assessed?hard
- A consultant works with eight clients. Each has its own contracts, tone of voice and history. Conversations must never mix one client's material into another's answers. Which arrangement best fits?hard
- A lawyer wants help structuring an argument. The underlying file contains client names and financial details. They would prefer this work not contribute to improving future models, and they are willing to change how they work but not their standing account settings. What is the soundest approach?hard
- An insurer uses Claude to draft the explanations that accompany claim decisions. A trained claims handler reads and approves every one before it goes out, and the letters make no reference to AI. Under Anthropic's Usage Policy, what is still owed?hard
- Someone asks Claude to "review our policy and flag anything risky", without attaching the policy, and gets back a generic list of things policies often get wrong. They conclude Claude is unreliable. Where does the fault actually lie?hard
- With web search on, Claude gives you a market-size figure and three citations. Following them: the first page states the figure outright, the second is a blog quoting an unnamed report, and the third no longer loads. You are about to publish the number. What is the sound judgement?hard
- You have ninety minutes to produce a briefing on how three competitors have priced a product over the past year, drawing on their sites, recent coverage and analyst commentary. You need something you can defend in a meeting. Which approach fits?hard
- Your usage ran out by lunchtime for the third day running. Nothing about your plan has changed. This week you started using Claude in your terminal alongside the web app, and you have been feeding whole 200-page manuals in rather than the extracts you used before. What is the most complete explanation?hard
Frequently asked
What people ask about claude.ai basics.
Is Claude free to use?
Why did I hit a usage limit sooner than I expected?
Does Claude remember what I told it in an earlier chat?
What is the difference between a project and a normal chat?
What are Artifacts, and why does my answer open in a side panel?
Can I trust what Claude tells me?
More Claude topics
Related guides
Limits and plans · 9 min read
What actually uses up your Claude usage limits
Anthropic publishes the plan tiers; it does not publish what a limit is spent on. The meter runs on tokens rather than messages, a long chat re-sends itself every turn, and attachments, project knowledge, tool results and thinking all land in the same budget. Mechanism first, numbers last.
How it works · 9 min read
Why the same prompt gives you a different answer
You changed nothing and the answer changed anyway. A mental model of what actually varies between two runs of an identical prompt: sampling that never fully stops, a context window that grows every turn, a system prompt the vendor rewrites, memory you cannot see, and infrastructure that moves under a pinned model.
Troubleshooting · 10 min read
Why ChatGPT and Claude didn't really read your file
The upload succeeded and the answer is about the wrong thing. Both vendors document the formats they accept; neither documents what happens between the upload and the model. Eight causes, each with a distinct fingerprint, and three probes that tell you what the model actually saw.