Deep Research PDF to HTML Article

Deep Research exports as PDF carry their citations in the document: inline superscript numbers, a trailing numbered sources section, and URI link annotations behind every citation. This tool decodes all of it. It reconstructs the text layout with a font-size-aware pass (body, section headings, display equations and citation superscripts have distinct, measurable sizes in these exports), reflows the extracted lines into paragraphs, and turns each citation superscript into a clickable link — resolving its URL directly from the PDF's embedded link annotation, with the trailing sources list (where one source can own several citation numbers) as fallback. Page-number footers and the ChatGPT watermark link are dropped; a table of contents is built from the section headings; output is one self-contained styled HTML file with a full references section. Honest limits: the PDF stores math as rendered glyphs (shown as centered text blocks, not re-typeset LaTeX) and figures/diagrams cannot be recovered — a note says so when figure-caption fonts are detected.