Published on August 12, 2026
Claude’s updated drafting experience lets users highlight specific passages in reports, plans and briefs, request targeted revisions and preserve the rest of the document without restarting the discussion.
Anthropic updated Claude’s drafting experience on June 12, 2026, making it easier to revise substantial written outputs without repeatedly explaining which section needs to change.
When Claude produces a report, plan, brief or another substantial Markdown document in Chat or Claude Cowork, the draft can open in a dedicated window beside the conversation. Users can highlight a specific passage, select Edit with Claude, and describe the revision they want. Claude then applies the change directly where the text was selected.
According to Anthropic’s Claude release notes, the purpose is to remove two common sources of friction: switching between Claude and a separate writing application, and trying to identify a paragraph through another long chat instruction. The feature turns Claude’s output into a more focused editing workspace rather than treating every draft as a static response.
Claude has long been able to produce lengthy written material, but revising that material through conversation could become inefficient.
A user might receive a multi-page report and then ask Claude to rewrite the third paragraph under a particular heading. If the document contained similar sections, the instruction could be misunderstood. A broader revision request might also cause Claude to regenerate more of the document than necessary, potentially changing wording that the user already wanted to keep.
The updated experience separates the draft from the conversational thread while keeping the two connected. The document appears in an artifact window to the right of the chat, where it can remain visible as the user discusses changes with Claude.
Anthropic defines artifacts as substantial, self-contained outputs that users are likely to edit, reuse or reference later. They can include text documents, code, diagrams, websites and interactive components. For written work, the side-by-side format gives the draft a stable location while the conversation continues alongside it. Anthropic’s artifacts documentation explains that changes appear directly in this separate window rather than as another disconnected version buried in the chat history.
The most important improvement is precision.
Instead of describing the location of an unwanted sentence, users can select it directly and ask Claude to make it shorter, clearer, more formal or better supported. The request remains tied to the highlighted passage, reducing ambiguity about what should be revised.
A professional reviewing a strategy brief could select an overly broad recommendation and ask Claude to make it more specific. A manager could highlight a section of a project plan and request clearer responsibilities and deadlines. A student could mark a confusing paragraph in a report and ask for a more accessible explanation without asking Claude to rewrite the entire assignment.
This editing model is particularly useful when most of a draft is already acceptable. The user can preserve its overall structure while concentrating attention on weak or incomplete sections.
The workflow also changes the role of prompting. A revision request no longer needs to repeat the heading, quote the original paragraph and explain where it appears. The selected text provides the reference, allowing the written instruction to focus on the desired improvement.
AI-generated drafts rarely move directly from first output to final publication. They often require changes to tone, evidence, structure, emphasis and wording.
The updated drafting experience supports this iterative process without forcing users to regenerate the complete document after every adjustment. Claude can revise one passage while leaving the surrounding sections intact, after which the user can select another area and continue refining the work.
Users can also move between previous artifact versions through a version selector. This provides a way to compare revisions or return to an earlier draft when a later change is less effective. Anthropic notes that modifying the artifact does not change Claude’s memory of the original content, so the conversation and document version are related but not identical records.
For documents spread across several Markdown files, users can leave editing requests in multiple files and submit them together. Claude then applies the batch of revisions in one pass. This may be useful for connected materials such as a project package, plugin documentation or a set of related internal instructions.
The result is closer to collaborative editing than ordinary prompt-and-response generation. Claude creates the initial material, while the user directs increasingly precise improvements.
The feature is most relevant to documents that are developed through several rounds of review.
Reports often require factual corrections, stronger conclusions and clearer explanations. Plans may need updated milestones or responsibilities. Business briefs can require changes to audience, tone or strategic emphasis after stakeholder feedback.
With side-by-side editing, a user can keep the source conversation visible while revising the deliverable. That is useful when the chat contains instructions, research context or earlier decisions that still inform the document.
The experience also fits naturally within Claude Cowork. Cowork is designed for longer tasks that may involve research, files, tools and professional outputs. Once Claude has completed the broader assignment, the resulting Markdown draft can be refined through targeted selections rather than sending the whole task back through another production cycle. Anthropic includes in-place draft editing among Cowork’s principal capabilities.
This could reduce unnecessary context switching. Users may still export a final document to Word, Google Docs or another platform for advanced formatting and collaboration, but more of the early drafting and revision process can remain inside Claude.
The update makes Claude more useful as an editing environment, but it does not turn the artifact window into a complete word processor.
In-place AI revision is currently documented for Markdown drafts. Complex page layouts, detailed styles, tracked changes, comments, citation management and formal approval workflows may still require dedicated document applications.
The quality of a revision also depends on the instruction and Claude’s interpretation. A precisely selected passage can still be changed in an unsuitable way, introduce an unsupported claim or become inconsistent with another part of the document.
Users therefore need to review each material revision and check the full draft for coherence. This is especially important for academic work, financial reports, legal documents, public communications and other outputs where accuracy has consequences.
The feature improves control over where Claude edits. It does not guarantee that every edit is correct.
Artifacts are supported across Free, Pro, Max, Team and Enterprise plans. Anthropic states that code execution and file creation must be enabled in the account’s capabilities settings—or by an organization administrator—for artifacts to function.
The in-place revision experience applies specifically to Markdown documents. When an eligible draft opens beside the conversation, users can highlight text and choose Edit with Claude to submit the targeted request. Artifact capabilities are available through Claude and Claude Desktop, while related artifact functionality is also supported in Claude Code for eligible organizational plans.
Exact interface details may evolve as Anthropic continues updating Claude, Cowork and the shared artifacts experience.
Claude’s updated drafting experience addresses a small but important weakness in conversational writing tools: revising a document should not require repeatedly describing where every change belongs.
By placing drafts beside the conversation and connecting revision requests to highlighted text, Claude allows users to refine individual sections while preserving the rest of the work. Reports, plans and briefs can evolve through targeted edits rather than complete regeneration.
The broader significance is a shift from AI as a text generator toward AI as a collaborative editing partner. The first draft remains useful, but the more valuable capability may be the ability to improve that draft precisely, transparently and over several rounds.
For professional productivity, this type of interaction can make AI-assisted writing feel less like requesting a succession of answers and more like working continuously on one document.
Anthropic Help Center — Claude Release Notes — In-place draft editing announced June 12, 2026.
Anthropic Help Center — What Are Artifacts and How Do I Use Them? — Updated July 2026; accessed August 5, 2026.
Anthropic Help Center — Get Started With Claude Cowork — Updated August 2026; accessed August 5, 2026.
