The reading-order problem in step 1 is the part people underestimate: a PDF that looks perfectly normal to a human eye can extract as a scrambled sequence of text fragments, and pasting that scrambled sequence straight into blocks just moves the mess from the PDF into the page instead of fixing it.
Gutenberg
Convert a PDF brochure's layout into structured block sections
A scenario prompt that rebuilds a product one-pager PDF as an ordered sequence of Gutenberg sections, since PDF text position doesn't map cleanly to reading order or heading hierarchy.
Works with Claude / GPT310 uses★ 4
The prompt
gutenberg-pdf-brochure-to-block-sections
Sales gave me a one-page PDF brochure for our new product line and wants its content turned into a page on the site. Here's the text extracted from the PDF, in the order it was pulled (not necessarily reading order, since PDFs often store text by position on the page rather than logical flow): [paste the raw extracted PDF text here, including any large font-size headings, a two-column feature list, a pull-quote, and a small-print disclaimer at the bottom] Rebuild this as a logically ordered page using core blocks, don't just dump the extracted text top to bottom in whatever order it was pulled: 1. First, reconstruct the actual reading order a human would follow on the printed page (usually top-to-bottom, left column before right column within the same row), since PDF text extraction frequently interleaves a two-column layout into one flat sequence that reads left-line, right-line, left-line, out of order. Call out clearly where you had to guess at intended order. 2. Identify real headings versus large decorative text: PDFs distinguish headings purely by font size and boldness with no semantic tag underneath, so infer the hierarchy from context (a product name is probably the H1-equivalent, a feature category label is probably an H2, an bolded price is not a heading at all) rather than making everything in a big font an `wp:heading`. 3. Convert the two-column feature list into a `wp:columns` block with matching `wp:list` content in each column, and the pull-quote into a `wp:quote` block. 4. Any image referenced in the PDF (a product photo, a diagram) will need to be re-exported at web resolution separately, since PDF-embedded images are usually print-resolution and far larger than a page needs, flag each one rather than inlining a placeholder and hoping someone remembers. 5. Put the small-print disclaimer at the very bottom in a `wp:paragraph` using a smaller font size setting, not the same size as body text, so it visually reads as fine print the way it did in the original PDF. 6. Output the full block markup in corrected reading order, with a short list underneath noting every place you had to make a judgment call about order or hierarchy.