What a copied Claude answer contains
- Writes in markdown: **bold**, ## headings, - lists two or three levels deep, fenced code with a language tag, pipe tables.
- The copy button under an answer gives markdown text; the copy button inside an Artifact gives the artifact's raw source (markdown, HTML or code).
- Numbered lists restart at 1 after a paragraph break, so word processors merge or renumber them.
- Uses spaced en dashes (–) and em dashes (—) freely; both read as the AI tell.
- Opens with a short preamble ("Here's a version that…") that rarely belongs in the document.
- Numbers and units can carry a narrow no-break space ("3 PM", "5 km") that word processors refuse to wrap on.
What is different about Claude's output
Every assistant writes markdown; Claude's is longer, more nested and a little more typographic than most. A Claude answer to a planning question is a heading per section, a numbered list of steps, bulleted details under each step and sometimes a third level under those. It uses spaced en dashes – like this – as well as em dashes, curly quotes, and a short preamble and sign-off around the content. Longer pieces go into an Artifact, a side panel with its own copy button that copies the artifact's raw source rather than the rendered text. When Claude has searched the web, the answer carries reference links and occasionally a sources list.
None of that is wrong in the chat window. All of it becomes a clean-up job the moment the text goes somewhere else, because the destination either shows the markdown literally (Word, Docs, email, Upwork, LinkedIn) or needs the structure rebuilt (a spreadsheet, a note app). The pages on this site split into those two jobs.
Removing the formatting, or converting it
The editor above removes. Paste a Claude answer and the markdown symbols go, headings become plain lines, links become readable text with the address, hidden characters are stripped, and bold and bullets are optionally kept as Unicode characters that plain-text boxes cannot strip. That is the right tool for LinkedIn posts, Upwork proposals, plain-text email, CRM notes and anywhere else formatting does not exist. The preferences add the Claude-specific fixes: replace the dashes with commas or hyphens, straighten the curly quotes, strip the rest of the markdown — italics, code, quotes, tables — and remove emoji.
The destination pages convert. Claude to Word and Claude to Google Docs turn the same markdown into real headings, bold, italic and nested lists, which is what you want inside a document. Claude to Excel and Claude to Google Sheets take Claude's tables — markdown, CSV artifacts, ASCII grids — and make real cells with the numbers typed. For Gmail, Outlook, Slack, Discord, Reddit, WhatsApp, Telegram and Notion, the platform formatters linked above read Claude's markdown the same way and write each destination's native format.
Artifacts, files and the copy buttons
There are two copy buttons in Claude and they do different things. The one under a chat answer copies the markdown of the answer. The one inside an Artifact copies the artifact's source — markdown for a document, HTML for a page, code for code — which is what you want for editing and what this site reads directly. Claude can also hand you a finished file (a .docx, a .xlsx, a .pdf) when you ask for one, and for a whole deliverable that is the shortest path. The pages here are for the far more common case: one answer, one paragraph, one table, going into something you are already writing.
The tell, and what to do about it
Readers have learned to spot generated text, and Claude's version of the tell is specific: the spaced dash, the evenly weighted paragraphs, the "I'd suggest" opener, the closing offer to revise. The tools fix the punctuation in one paste; the em dash remover explains the comma-versus-hyphen choice. The preamble and sign-off you delete by hand, and the structure you loosen by editing — shorter paragraphs where the point is simple, a longer one where it is not. Formatting makes the text look typed; editing makes it read written.