PDF CATALOGS AND HTML PRODUCT PAGES: FIVE PAIRED EXAMPLES Neaptide original research, observed October 4, 2026 Study ID: PDF-PAIRS-2026-10-04 QUESTION AND SCOPE Which product information appears in HTML and in a linked PDF covering the same model or explicitly qualified family? The unit is one verified pair per company. This is a content and navigation comparison, not a test of SEO performance. SELECTION BEFORE WRITING We reused the fixed B2B-36 parent sample: 36 product URLs across 12 companies serving B2B buyers. That sample is a convenience sample, not representative of US manufacturers or distributors. This follow-up is not an independent sample. The protocol and ordered candidate list were saved and hashed before PDF collection. We checked the first document candidate for each company's first eligible page, then moved through the registered candidates if a match could not be verified. No new companies or external PDF searches were used to fill the sample. Five companies yielded a pair: Pacron, Carr Lane Manufacturing, norelem, USA Lab and Galco. Five companies had no product-specific document candidate in their three parent-study pages. Flow-Eze's two recorded PDF links led to the same general technical reference without the vial identifiers. Four LabRepCo download URLs returned HTTP 403 to our collector; we did not infer that normal visitors could not access them. Pacron's first candidate, Bubble Roll, was not accepted: HTML and PDF variant codes differed and an explicit mapping was not verified. The next candidate, Foam Roll, contained the same F13248 identifier in both formats. Five is below the protocol's six-company editorial threshold for a field-study label. We therefore present examples. The threshold is not statistical significance. WHAT WAS MATCHED Pacron: F13248, a specific foam-roll variant. Carr Lane: CL-1-LP-ZC, with family-level material/cable information. norelem: 01040-01, within product family 01040. USA Lab: base centrifuge model UGL20 / USA-UGL20, not every bundled rotor kit or firmware version. Galco/IMO: explicitly qualified Series 69 / PE69 family coverage. The PDF's ordering-code key and family table were checked. The full PE69-3020 SKU is not printed as one continuous text string in the PDF; individual enclosure dimensions were not proven equivalent. COLLECTION AND ACCESS We used the browser to inspect rendered HTML and ordinary specification, description, document and material/cable sections. We saved dated private DOM/text captures. We did not log in, submit a lead or purchase anything. Public linked PDF files were downloaded with a bounded Python HTTP collector. We saved response status, content type, capture time, byte size and SHA-256. One malformed decoded Galco URL failed locally before a request; its record was retained and spaces were correctly URL-encoded for the successful request. The Carr Lane viewer initially timed out or showed a loading error, then loaded normally. We returned to the first candidate, catalog page 32. The site's Download PDF > Page Range selector exported only page 32 without a cover. We clarified the protocol's form restriction before this read-only export: it prohibits inquiry/account/purchase submissions, while a public document page selector is permitted. The generated PDF URL is temporary; use the stable catalog-page link and repeat the native export. Generated bytes may differ on a later export. TEXT AND VISUAL CHECKS Text extraction used pypdf 6.10.0. PDF pages were rendered with Poppler and inspected visually. Each selected file was at most 17 pages; the protocol allowed up to 20 reviewed pages per candidate. The selected PDFs contain 25 pages in total, including a one-page Carr Lane excerpt and the two-page norelem file. Four of the five selected PDFs yielded extractable text. Pacron's one-page PDF contains a page image and yielded zero text characters to pypdf; we read the rendered page. This is not proof that Google or another engine cannot recognize or index that content. No OCR system, Google index or AI answer engine was tested. The visible Carr Lane footer reads 8/26. Extraction also returned an older 8/19 string; we report the visible revision mark. The IMO datasheet displays reference 1121 v8. File metadata dates are not treated as a supplier's verified publication or revision date. CODING Each pair has six tasks in each format, producing 60 field-format observations: 1. Model or part/family identifier. 2. Physical dimensions with units for the product, variant or relevant component. 3. Named construction or packaging material. 4. Explicit compatible component, accessory, interface or standard; record family-level scope. 5. Access to the selected technical document. 6. A route to buy, request a quote, contact or obtain support; distinguish working HTML links from printed PDF contact details. States are observed, not_observed, and unclear_units. Numeric dimensions with no explicit unit in the reviewed view are unclear_units, not absent. A generic use case or a related-product card does not prove compatibility. A homepage URL or a direction to 'contact us' without contact details is recorded in the notes but does not count as a direct contact/purchase route. Carr Lane's order-code instructions are likewise distinguished from a purchase destination. Printed Pacron phone/email details count as a contact route, with manual entry noted. The document-access task is structural: the HTML links to the PDF and the PDF is the inspected document. Do not use it in a quality score. We do not sum heterogeneous tasks into a winner or claim that a shared field has equal detail in both formats. The weight/load columns and the USA Lab 10-versus-100 program descriptions are exploratory observations, not preregistered population metrics. The correct program count, hardware/firmware revision and supplier specification accuracy remain unresolved. RECHECK The same researcher rechecked the HTML evidence for all five pairs and the relevant saved PDF pages. We compared exact model strings, key values, links and apparent differences. This is not independent coding or a second temporal PDF collection. The detailed dataset includes source links, HTML sections, PDF page references, capture dates and hashes. For Carr Lane, PDF page 1 corresponds to printed catalog page 32. LIMITS AND REUSE These five cases do not estimate market prevalence. Not-observed refers to the selected file or reviewed page state, not an entire supplier website. We did not test physical compatibility, equipment operation, delivery, rankings, indexing, traffic, citations, conversions or user task speed. Prices were not studied; Galco's observed browser session displayed TRY/TR settings. Public downloads contain our coded observations, methods and analysis. They do not redistribute supplier PDFs, complete page text, response headers or session details. Follow the source links to inspect the documents. A later page may differ from the captured version; the hashes identify the versions used here. TECHNICAL CONTEXT Google lists PDF among supported indexable file types: https://developers.google.com/search/docs/crawling-indexing/indexable-file-types For genuinely duplicate documents, Google documents canonical HTTP Link headers, including non-HTML files: https://developers.google.com/search/docs/crawling-indexing/consolidate-duplicate-urls Neither statement establishes the performance of the five examples. A complementary manual is not automatically a duplicate of a short product page.