Is the <think> block removed before it reaches Docs?
Yes. It is removed on paste into the editor, blank lines included, and only the answer is copied for Docs. Paste the reasoning separately if you want to keep it.
Paste a DeepSeek answer and copy it into Google Docs as native formatting: the <think> reasoning block removed, ### headings mapped to Docs styles, numbered lists kept with their nesting, markdown and hidden characters gone.
Paste the copied answer into the editor above. A <think> block in front is removed automatically, the markdown becomes visible structure, and hidden characters are stripped. Click Copy and paste into your Doc: headings become Heading 1–3, bold and lists stay, code keeps a monospaced font. DeepSeek's LaTeX stays as typed — use Insert → Equation in Docs or ask for plain-text maths. Nothing is uploaded.
Use the copy button under the answer or select the text. DeepSeek's copy gives markdown text, sometimes with the reasoning block in front; either form is read correctly here.
Markdown becomes real structure, hidden characters are removed, and the editor shows exactly what will be pasted.
Click Copy and paste into Google Docs. Headings, bold, lists and quotes arrive as native formatting.
Google Docs can paste markdown (Edit → Paste from Markdown), and for a plain DeepSeek answer it renders the ### headings and the numbered lists well enough. It has no idea what to do with a <think> block. The reasoning that DeepSeek's thinking models produce — several paragraphs of the model working through the problem, wrapped in <think> and </think> — arrives in the clipboard whenever you drag-select from above the answer, use the API, run the model locally through Ollama or LM Studio, or use a third-party client. Pasted into a Doc, it sits at the top of the document as a wall of first-person deliberation.
This page removes the block on paste, with the blank lines it leaves, and shows only the answer in the editor. It then does what Paste from Markdown does not: strips the hidden characters, lets you see and adjust the structure before it lands, and copies rich text that Docs pastes as its own headings and lists without needing the markdown setting switched on.
DeepSeek likes ### as a top-level heading and long numbered lists whose items start with a bold phrase. The converter maps the headings it finds to Heading styles — a ### becomes Heading 3; promote it in the editor if the document outline should start at Heading 1 — and the numbered lists become real Docs lists, nesting intact, with Tab and Shift+Tab working afterwards. Bold lead-ins stay bold. Fenced code blocks keep a monospaced font. Quotes become indented paragraphs. Tables are the exception: they do not travel on the rich copy, and the DeepSeek to Google Sheets page handles them, with the copied grid pasting into a Doc as a Docs table if that is where it belongs.
Watch for DeepSeek's occasional full-width punctuation — a wide comma or colon from the Chinese keyboard layout — in English answers; the editor shows it and you fix it before copying. The em dashes and colons that chain its clauses are kept as typed, and the em dash remover replaces them if the Doc should not read as generated.
DeepSeek writes maths as LaTeX — \( E = mc^2 \), $\frac{a}{b}$ — whenever a question is quantitative. Docs shows it as typed, and the converter does not guess a rendering. Docs' own equation editor (Insert → Equation) accepts a subset of LaTeX commands typed into the equation box, which handles fractions, superscripts and Greek letters well; for anything heavier, the Auto-LaTeX Equations add-on renders $…$ expressions in place. The simplest route for a document with a lot of maths is to ask DeepSeek for "plain-text maths, no LaTeX" and let it write x² and a/b from the start.
Heading 1–3 styles, real bold and italic, numbered and bulleted lists with nesting, indented quotes, monospaced code, clickable links — and nothing else: no reasoning block, no hidden characters, no background colour. The pasted text inherits the Doc's font, and comments, suggestions and version history work on it as on anything typed. The conversion runs in your browser; the answer is never uploaded.
Yes. It is removed on paste into the editor, blank lines included, and only the answer is copied for Docs. Paste the reasoning separately if you want to keep it.
It renders the headings and lists but leaves a <think> block, hidden characters and full-width punctuation untouched. This page handles those and shows the structure before it lands.
It is kept as typed. Use Docs' Insert → Equation for individual expressions, an add-on such as Auto-LaTeX Equations for many, or ask DeepSeek for plain-text maths.
Via DeepSeek to Google Sheets: copy the grid there and paste it into the Doc, which turns it into a Docs table.
No. The conversion runs in your browser with client-side JavaScript. The DeepSeek answer is never sent to a server, stored or logged.