Why it matters
A pump distributor moves its Goulds 3196 performance curves and materials-of-construction tables out of a downloadable PDF and into HTML spec tables on the product page. Soon after, it starts getting cited for "Goulds 3196 max temperature" queries in AI answers. The data did not change. Its location did. A PDF is a closed file an engine has to open, parse, and trust; an HTML table is text the crawler reads as part of the page. Most datasheet value sits in the second column of a row, the bore size, the pressure rating, the seal material. Lock that in a PDF and the engine cannot lift the line that answers the buyer.
HTML tables vs PDF datasheets
Both formats hold the same numbers. They are not equal to a retrieval engine.
- HTML spec table: each spec is a labeled row, indexed with the page, easy to quote as a passage.
- PDF datasheet: a separate file, often image-scanned, frequently uncrawled, hard to attribute back to your URL.
- Keep the PDF as a download for engineers. Mirror its contents as HTML so the spec is also a passage on the page.
In practice
Take a hydraulic cylinder SKU. Build one HTML table with attribute names in the left column and values in the right: bore, rod diameter, stroke, max pressure, mount style, port size. Use real labels, not vague headers. Now the line "Max pressure: 3,000 psi" reads as a clean attribute-value pair an engine can cite verbatim. Pull the same normalized attributes from your PIM so every product page across the catalog renders specs the same way.