- In short
- The output format taxonomy names three main categories: inline for conversational responses acted on within the chat, artifacts for standalone documents or code that will be refined and reused as a separate editable deliverable, and structured formats such as tables or defined schemas for data meant to be consumed directly by tools or readers. Choosing among the three is a deliberate decision tied to what the output is for, not a default setting.
Three homes for an output
Where an output lives shapes how it can be used. The Claude Certified Associate - Foundations (CCAO-F) exam starts task statement 2.6 by naming three output format categories, each suited to a different purpose: inline, artifacts, and structured formats. Knowing what each is for is the foundation for treating format as a deliberate choice rather than accepting whatever shape a response happens to take.
The three categories are not interchangeable styles. They correspond to genuinely different uses, a reply you act on in the moment, a deliverable you refine over time, and data another system or reader consumes directly. Matching the category to the use is the skill this knowledge point establishes.
- Output format taxonomy
- Three main output format categories. Inline: conversational responses acted on within the chat, quick and contextual, not a standalone artifact. Artifacts: documents and code that will be refined and reused as a separate, editable deliverable. Structured formats: tables and defined schemas for data meant for downstream tools or readers to consume directly. Choosing among them is a deliberate decision tied to what the output is for.
Inline output
Inline output is the conversational response, delivered in the flow of the chat and acted on there. It suits quick, contextual answers, a question answered, a suggestion offered, a point clarified, where the output is not meant to become a standalone artifact. Inline is the default shape of a conversation, and it is the right choice when the answer's job is to be read and used immediately, in context, rather than to be extracted, refined, or handed off. Its strength is speed and immediacy; its limit is that it is not built to be reused as a separate deliverable.
Artifacts
Artifacts are for documents and code: a separate, editable block you refine and reuse, distinct from the conversational reply. When the output is a deliverable rather than a remark, something you will iterate on, come back to, or hand to someone else, an artifact is its proper home. It gives the output an existence outside the message stream, so it can be edited and developed without being buried in the conversation. Choosing an artifact signals that the output is a thing to be worked on, not just a reply to be read.
Structured formats
Structured formats, such as tables or defined schemas, are for data meant to be consumed directly by a downstream tool or a reader. When the output's purpose is to feed a system, populate a spreadsheet, or present data in a shape a reader can scan and use without restructuring, a structured format fits. The point is machine- or reader-readiness: the data arrives in the shape its consumer needs. Choosing prose when the consumer needs a table forces someone to restructure it, which is exactly what a structured format avoids.
What the CCAO-F exam trips candidates on
The first trap is defaulting to inline output for a deliverable meant to be reused, edited, or handed off. Inline is the conversational default, and letting it carry a document or code that should be an artifact leaves the deliverable trapped in the message stream. The credited answer chooses an artifact for anything meant to be refined and reused.
The second trap is requesting structured data output and then formatting the result as free-flowing prose instead. If the consumer needs a table or schema, prose defeats the purpose by forcing restructuring downstream. The exam rewards matching the delivered shape to the requested and required structure, not narrating data that was supposed to arrive structured.
Worked example
You ask Claude to produce a reusable project plan you will edit over several days, and separately to extract contract dates into something your tracking spreadsheet can ingest. Claude returns both as inline prose paragraphs. What format should each have taken, and why?
Neither output should have been inline prose, because inline is the conversational default and neither of these is a conversational response acted on in the moment. Match each to its purpose.
The reusable project plan is a deliverable you will edit over several days, which is the defining case for an artifact. It needs an existence outside the message stream so you can return to it, refine it, and develop it without digging it out of the conversation each time. Delivered as inline prose, it is trapped in the chat and awkward to iterate on, which is exactly the first trap: defaulting to inline for something meant to be reused and edited. It should be an artifact.
The contract-date extraction is data meant to be consumed directly by your tracking spreadsheet, which is the defining case for a structured format, a table or defined schema the spreadsheet can ingest. Delivered as free-flowing prose, it forces someone to restructure it before the spreadsheet can use it, defeating the purpose. This is the second trap: producing prose where structured data was needed. It should be a structured format matching what the downstream tool expects.
The general lesson is that format is a deliberate decision tied to what each output is for, a document to refine, data to feed a tool, not a default the conversation hands you. The project plan wants an artifact; the extraction wants structured data; neither wants inline prose.
Common misreadings to avoid
Misconception
Inline output is fine for any response, including deliverables.
What's actually true
Misconception
Structured data and prose are interchangeable ways to present the same information.
What's actually true
How this shows up on the exam
Domain 2 questions on this knowledge point present an output and ask which format category fits, or point out a mismatch between the intended use and the delivered shape. The dependable answer matches inline to conversational responses, artifacts to reusable documents and code, and structured formats to data consumed directly, treating the choice as deliberate rather than default.
This taxonomy is the foundation for the rest of task statement 2.6. It leads into output format selection as a reliability decision, the code execution versus prose generation choice for numbers, and the combined format and curation strategy for a task. It also connects to audience calibration, since format is part of fitting a deliverable to its reader.
You need a reusable project plan you will edit over several days, and separately a contract-date extraction your spreadsheet can ingest. Which formats fit?
People also ask
What are the main Claude output formats?
When should you use an artifact instead of inline?
What is structured output for?
Watch and learn
Official Anthropic Academy lessons first, then hand-picked walkthroughs. Videos load only when you press play.
No videos curated for this concept yet
We are still curating the best official and community videos for this topic.
Official prep for this domain
Anthropic's own free prep module for this part of the syllabus, on the official prep course. Free with an Anthropic Academy sign-in.
References & primary sources
Master this concept with Archie
Practice it inside an adaptive study session. Archie, your Socratic AI tutor, tracks your mastery with Bayesian Knowledge Tracing and schedules the perfect next review.