Choosing the right Claude feature

What you’ll be able to do

• Choose chat, a Project, an Artifact or research for a given business task and justify the choice
• Spot recurring work that should move into a Project with instructions and knowledge
• Decide when a multi-source research report is worth the extra time

What you need to know

Start from the shape of the work, not the feature list

Every feature question on the exam is really a question about the task. Before you think about features, ask three things: Will I do this again? (reuse), Will I keep editing the output? (iteration), and Does the answer depend on many outside sources? (evidence). The answers point to the feature almost every time.

• One-off, self-contained question or draft → a normal chat.
• Recurring work that needs the same background every time → a Project.
• A substantial piece of content you'll revise, reuse or share → an Artifact.
• A question that needs many sources pulled together with citations → research.

These combine. A marketing lead might work inside a "Q3 campaign" Project, ask for the campaign brief as an Artifact, and use research once to scan what competitors have announced. The skill being tested is choosing the right tool for each step, not picking one feature for everything.

Chat: the default for quick, self-contained work

A plain chat is the right choice when everything Claude needs fits in your message: rewrite this paragraph, suggest five subject lines, explain this spreadsheet formula, summarize these pasted meeting notes. It is fast and has no setup. The cost of chat shows up only when you find yourself repeating the same context over and over, or when the output grows into something you'll keep editing.

Projects: shared context for recurring work

A Project groups chats around a piece of ongoing work and gives them shared context: Project instructions (standing guidance such as audience, tone and format) and Project knowledge (files you upload, such as a style guide, a policy handbook or last quarter's reports). Every chat started inside the Project can draw on that context, so you don't have to re-explain it.

• Good fits: a weekly newsletter, an HR policy FAQ assistant, a client engagement, a course you teach every term.
• The context belongs to the Project. A chat started outside the Project doesn't see that Project's instructions or files.
• Depending on your plan and your organization's settings, Projects may be shareable with colleagues. That makes them a way to give a team consistent results, and also a reason to be careful about what you upload.

Configuring Projects well is covered in Domain 5. For this domain, the key judgment is recognizing when the work is recurring enough to justify one.

Artifacts: content you’ll iterate on or reuse

An Artifact is a standalone piece of content shown alongside the chat, such as a document, a table, a simple visual or a small interactive page. It suits output that is substantial and that you'll keep working on: a two-page policy draft, a project plan, a lesson handout, a comparison table. You can ask Claude to change specific parts and see the updated version, instead of hunting through a long scroll of chat replies for the latest draft.

Short answers don't need an Artifact. A one-line definition or a quick yes/no belongs inline in the chat. The question to ask is whether this will live on as a thing you edit, copy out or share.

Research: multi-source questions with citations

Research mode has Claude search across multiple sources and come back with a report that cites them. It takes longer than a normal reply and is meant for questions where the answer has to be assembled from many places: "What are the main approaches other universities take to AI-use policies for students?" or "Summarize recent public announcements from our three main competitors."

• Use it when breadth and sourcing matter more than speed.
• Don't use it for things Claude can answer directly, or for questions that are only about documents you have already provided.
• Citations make checking easier, but they are not a guarantee. Open the sources behind any claim you plan to rely on or pass along (Domain 2).

Web search is the lighter-weight option. Use it when you need one current fact or a quick check rather than a full report. Feature availability, including research, connectors and memory, varies by plan and admin settings, so the exam frames these as choices rather than as guaranteed options.

Key concept

Choose the feature by asking whether the work recurs (Project), will be iterated on (Artifact), needs many cited sources (research), or is a quick one-off (chat).

Exam scenarios usually describe a pain point, such as "re-uploading the same style guide every week", "losing track of the latest draft" or "needs a sourced overview of a market". The correct answer is the feature that removes that specific pain, not the most powerful-sounding one.

Practice scenario

Hands-on exercise — Map one week of your work to features

Beginner · 15 min

What you’ll learn

• Classifying tasks by reuse, iteration and evidence needs
• Seeing the difference between an inline reply and an Artifact
• Judging when research is worth the wait

1. List five tasks you did with Claude (or would like to) this week, and label each “one-off”, “recurring”, “will iterate” or “needs many sources”.

   • Why: The label, not the topic, decides the feature.
   • You should see: At least one recurring task that you’ve been re-explaining from scratch each time.

2. In a chat, ask Claude for a one-page document you’ll actually reuse (for example a meeting agenda template), then ask it to change one section.

   • Why: This shows how an Artifact keeps a single, current version you can edit and copy out.
   • You should see: The content shown as a standalone document beside the chat, updated in place when you ask for a change.

3. If research is available on your plan, ask the same multi-part industry question once in a normal chat and once with research. Compare the time taken, the depth and the citations.

   • Why: You’ll get a feel for when the slower, sourced report is worth it.
   • You should see: A longer, cited report from research. Open two of the citations and confirm they say what the report claims.

Exam traps

Pasting the same background documents into a new chat every week instead of using a Project

Repeated context is the clearest sign of recurring work. A Project holds instructions and files once for every chat inside it.

Using research mode for a quick question that a normal chat answers in seconds

Research is slower and meant for multi-source questions. For a definition, a rewrite or a quick fact, it only adds waiting time.

Keeping a long document you’ll keep revising buried in chat replies instead of an Artifact

Substantial content you’ll edit, reuse or share is easier to manage as an Artifact, which keeps one current version next to the chat.

Assuming a Project shares its knowledge with chats outside the Project

Project instructions and files apply to chats inside that Project. A chat started elsewhere doesn’t see them.

Treating a research report’s citations as verified just because they are there

A citation shows you where to check, but it doesn’t prove the claim. Open the sources for anything you’ll rely on.

Picking the feature with the most power rather than the one that fits the task

The exam rewards fit. Look for the option that removes the specific pain in the scenario.

Sources

• Claude Help Center
• Claude Docs
• Anthropic, Claude Certified Associate – Foundations Exam Guide v1.0 (July 2026)