Claude is Anthropic's AI assistant — available at claude.ai and in the Claude app. This guide is for people using Claude through the consumer interface rather than the API. It covers everything Claude can do in practice: long documents, artefacts, projects, extended thinking, and the features that make Claude distinctively useful for professional and creative work.
Claude is an AI assistant made by Anthropic. You talk to it at claude.ai or in the Claude mobile app. It reads and writes text, understands and creates code, analyses documents, and helps with an enormous range of tasks — from drafting emails to reviewing contracts to writing software.
Claude is particularly known for three things: handling very long documents (it can read a full book or a large codebase in one conversation), following complex instructions carefully and consistently, and being genuinely useful for professional work rather than just answering questions.
Claude vs ChatGPT — the practical difference: Both are capable AI assistants. Claude tends to give more nuanced, careful responses and excels at long-context work. ChatGPT has a larger plugin ecosystem and is more widely known. Many professionals use both for different tasks.
Reading and analysing documents — contracts, research papers, reports, datasets. Paste the document and ask questions, request summaries, or ask Claude to identify specific information.
Writing that requires careful instruction-following — style guides, specific tone requirements, structured formats, templates. Claude tends to follow detailed instructions more consistently than most alternatives.
Code review and explanation — paste code and ask Claude to explain it, find bugs, suggest improvements, or refactor it to meet specific requirements.
Extended professional work — multi-step projects like research synthesis, drafting documents with multiple sections, or iterating on complex creative work benefit from Projects and the long context window.
Projects are the most underused feature in Claude. Create a project for each ongoing area of work (client, topic, or project type). In the project instructions, tell Claude: your role and context, any preferences or constraints, recurring tasks, and how you want it to respond. This context persists across every conversation in the project — no need to re-explain yourself each time.
Example project instructions: "I am a marketing manager at [company type]. I work on B2B content — blog posts, case studies, and email campaigns. Always use active voice. Avoid marketing jargon. When I ask for drafts, give me one version, not multiple options, unless I ask."
Claude is available in three model tiers:
On claude.ai, the Free tier uses Claude 3.5 Sonnet with usage limits. Pro unlocks higher usage and access to Opus. The specific models available update as new versions are released — check anthropic.com/claude for the current model lineup.
Anthropic trains Claude using a method called Constitutional AI (CAI), published in a 2022 research paper (Bai et al., arXiv:2212.08073). CAI involves training the model to evaluate and revise its own outputs against a set of principles (the "constitution") — covering helpfulness, harmlessness, and honesty — rather than relying solely on human feedback for every behaviour. This approach is designed to produce models that are more reliably aligned with intended values at scale. The full Anthropic model card and Constitutional AI paper are publicly available at anthropic.com/research.
Claude 3.5 Sonnet has a 200,000-token context window — approximately 150,000 words or roughly 500 pages of text. This is one of the largest context windows available in a consumer AI product and is the basis for Claude's document analysis and long-context capabilities. For comparison, a typical novel is 70,000-100,000 words; an entire codebase for a medium-sized application typically falls within this window.
By default on claude.ai, conversations may be used by Anthropic to improve models. Users can opt out of this in Settings. On Claude Team plans, conversations are not used for model training by default. Full privacy policy at anthropic.com/legal/privacy.
Source note: Pricing from claude.ai/upgrade. Technical specifications from Anthropic model documentation. Constitutional AI from Bai et al. (2022), arXiv:2212.08073. All verified April 2026.