Converts most web page elements into clean Markdown with visual element selection and live editor preview support. This Trusted-Types safe script supports extraction and conversion from ChatGPT deep research overlays with sub-element selection inside the maximized panel, including citation/sources/thinking-activity extraction, and Google Gemini deep research reports and canvases, featuring 6 citation styles, YAML frontmatter integration, and exporting to clipboard/file/GitHub/Obsidian.
Converts highlighted html elements from most web pages into clean Markdown, expressly integrating conversion and export support for ChatGPT Deep Research reports and similar elements across various web pages, including citation extraction with source URLs from sandboxed iframes and srcdoc-embedded report handling, featuring 6 citation styles, YAML frontmatter support, a visual element picker, live preview editor, and clipboard/file/GitHub/Obsidian export capability.
ChatGPT's Deep Research reports (as of ~March, 2026) are rendered inside sandboxed <iframe> elements on web-sandbox.oaiusercontent.com. Standard web clippers and HTML-to-Markdown tools cannot access these iframes, causing userscript malfunctions and empty or malformed outputs. Saved/exported ChatGPT pages embed the report as HTML-entity-encoded srcdoc attributes with multiple layers of encoding. Citation URLs are stored in React's internal fiber state, invisible in the static HTML. This script rectifys all three challenges, building upon the exceptional work of shiquda and ChinaGodMan's MarkDown Cloud Cut Notes.
This tool is a direct fork of shiquda and ChinaGodMan's MarkDown Cloud Cut Notes (v2025.03.19), with additional inspiration drawn from Chris Kephart's exemplary ChatGPT Deep Research Markdown Exporter. The source code has been substantially refactored, introducing the following revisions:
postMessage to communicate with the sandboxed iframe, extracting the full report content including headings, paragraphs, tables, code blocks, and citations__reactFiber internal properties on <sup data-citation-index> elements to recover source URLs from props.item.reference.safe_urls<iframe srcdoc="..."> attributes([1](url)), inline [1](url), endnotes [1], footnotes [^1], named ([domain](url)), or none — with automatic deduplication<mark> highlights, <details> blocks, task list checkboxes, iframe skipCtrl+M (or use the Tampermonkey menu → "Convert to Markdown") to enter element selection mode. Use arrow keys or scroll wheel and cursor to navigate across elements, then click to convert the highlighted element of interest.R during selection mode, or use the Tampermonkey menu → "Export Deep Research" to auto-detect and extract the full report with citations..html file in your browser. The script detects srcdoc iframes and extracts the embedded report..md file| Key | Action |
|---|---|
Ctrl+M |
Start element selection |
| Arrow keys | Navigate elements (parent/child/sibling) |
| Scroll wheel | Expand/shrink selection |
R |
Auto-export Deep Research content |
Esc |
Cancel selection or close modal |
| Click | Convert selected element to Markdown |