What a copied Grok answer contains
- Markdown with **bold**, ### headings, pipe tables and fenced code; the copy button gives markdown text.
- When it searches X or the web, numbered source markers and link lists ride along with the answer.
- Casual register: emoji, dashes and rhetorical asides that read as generated in a formal document.
- Research answers run long, with nested lists and several tables.
- Maths as LaTeX when the question is quantitative.
What is different about Grok's output
Grok is xAI's assistant and lives inside X as well as in its own app, which shapes its output in two ways. Its sources are often X posts: a searched answer carries numbered markers after the sentences it drew from them and ends with a list of the posts and pages used. And its voice is conversational by design — asides, dashes, emoji, a closing quip — which is fun in a chat and a liability in a proposal. Underneath, the markdown is standard: **bold**, ### headings, hyphen bullets, pipe tables, fenced code, LaTeX for maths. The copy button hands you all of it as text.
A lot of Grok output also goes straight back to X. That is its own formatting problem: X has no bold, counts Unicode bold and emoji as two characters each, and shows markdown literally. The editor above is built for exactly that round trip.
Removing the formatting, or converting it
The editor above removes. With Remove citations on — the default on this page — the inline source markers go. The markdown goes with them: headings become plain lines, links become readable text with the address, hidden characters from quoted posts and pages are stripped, and bold and bullets can be kept as Unicode characters that plain-text boxes cannot strip. The preferences add the rest: Remove emojis, which Grok answers need more than most, the em dash replacement, and the full markdown strip. For posting to X, the X text formatter adds the counter that weighs characters the way X does.
The destination pages convert. Grok to Word and Grok to Google Docs remove the markers, keep the source list as links, and turn the markdown into real headings and lists. For Grok's tables, the ChatGPT table to Excel and Google Sheets converters read the same markdown pipes and type the numbers. For Gmail, Outlook, Slack, Discord, Reddit, WhatsApp, Telegram and Notion, the platform formatters linked above write each destination's native format.
Citing X posts
Removing a marker does not remove the source, and when the source is an individual's post the document should say so. Every Grok page here keeps the source list as clickable links; turn it into footnotes in Word or Docs, a references block, or a Source column in a sheet, and delete the list once your citations are in. For a claim whose provenance matters, ask Grok to name the source in the sentence ("as @account reported on 3 March") rather than relying on a marker that will be stripped — and verify the post, because an answer built on social posts inherits their reliability.
The register, and what to do about it
Grok's version of the generated-text tell is tonal rather than structural: the aside in dashes, the emoji, the quip at the end, the "Let's dive in". The tools fix the punctuation and the markup in one paste — the em dash remover handles the dashes, Remove emojis the rest — and the voice you fix by cutting: the opener, the asides, the closing line. Asking Grok for a formal or concise tone at the start, and avoiding the Fun mode where offered, leaves far less to cut. Formatting makes the text look typed; editing makes it read written.