Differentiated the Chinese (and ten-locale) batch webpage-to-PDF vs webpage-to-JPG pages so each has independent information gain: A4 pagination/print archive versus viewport JPEG screenshots.

Date: 2026-09-07 13:20 Summary: Differentiated the Chinese (and ten-locale) batch webpage-to-PDF vs webpage-to-JPG pages so each has independent information gain: A4 pagination/print archive versus viewport JPEG screenshots. Visibility: people

[question] https://onlinefreetools.org/zh/tools/batch-convert-web-pages-to-jpg 和 https://onlinefreetools.org/zh/tools/batch-convert-web-pages-to-pdf, 页面内容过于相似，info gain不足

Tool links (English)

[try to solve]

Context

The two tools share a URL-list fetch pipeline (proxy once, skip failures, ZIP). After launch, the visible How / Rules / FAQ / lead still started from the same “paste URLs → Convert all → ZIP / fetched once, not stored” skeleton. Removing the H1 left two pages that read as format-suffix variants, which is a doorway / scaled-content risk under Google’s helpful-content and spam policies (information gain is a people-first content principle, not a claimed ranking switch).

The actual products already differed: PDF uses html2pdf with A4 portrait and CSS page-break; JPG uses html2canvas with viewport 1280/768/390, JPEG quality, and full page vs first screen.

Process

Compared zh/en shards, catalog related, and Page sample HTML. JPG already named viewport/quality; PDF barely mentioned A4 and never FAQ-compared the JPG sibling.
Rewrote IG-bearing copy in all ten locales independently (not an English calque): PDF lead/meta/How/Rules/FAQ/Use cases now open on printable A4 pagination; JPG opens on screenshot width, chat/ticket rasters, and first-screen review.
PDF FAQ question order starts with the JPG contrast (or paper size, depending on locale) instead of the shared upload FAQ. Added a sixth FAQ on A4 vs one tall page.
PDF catalog related now includes batch-convert-web-pages-to-jpg. PDF hero uses the short desc (same pattern as JPG) so the first screen is not a cloned meta paragraph.
Sample pages no longer share the same Georgia help/policy HTML: PDF samples use @page A4 plus page-break-after; JPG samples use a system-ui ticket header that turns blue at max-width 500px. JPG sample URLs are ticket/mobile-hero; PDF keeps help/policy.
Raised PDF howItemCount to 6 and ruleItemCount to 5 so the merge-neighbor How step and the JPG/HTML/combine Rules contrast actually render.
Ran coverage:gate --phase=all on both slugs, merge:tools, full build-site, HTML smoke, lint:seo (OK), isolation with CROSS_TOOL_UPDATE=1.

Root cause / analysis

Shared input (URL list + ZIP) is real, but that is not enough unique copy. The missing IG was the artifact contract:

PDF: A4 portrait, CSS page breaks, print CSS can move cuts, archive/read/print, then Combine files if you want one booklet.
JPG: chosen viewport layout, JPEG quality, full-page vs first-screen raster for chat and QA.

Without those sentences in How/Rules/FAQ, crawlers and people only see “batch convert web pages to {format}”. That fails the “strip the title, is the body still different?” test in tool-i18n-seo.mdc.

Honest boundary: html2pdf still captures via canvas then slices onto A4. Copy does not claim selectable text; it claims pagination and print-oriented pages.

Solution

After this pass, stripping H1 still leaves two jobs:

PDF zh lead: “把网址列表存成可打印的 A4 纵向 PDF…不是网页长图.” FAQ 1 vs JPG; Rules name CSS page-break; related includes the JPG tool.
JPG zh lead: “按桌面/平板/手机视口把网址列表截成 JPEG…不是 A4 PDF.” Example uses ticket + mobile-hero samples; Advanced still documents viewport/quality/first screen.

Ten locales keep local H1s and different FAQ order so they are not isomorphic translations of one English block.

Notes / boundaries

Isolation for a single slug needs CROSS_TOOL_UPDATE=1 because both tools were edited in one pass.
JPG still does not auto-run loadSample() (html2canvas can freeze the tab); PDF same exception.
Production Chinese URLs will show this after the next Pages deploy; local prerender already contains the new How sixth step and related JPG link.

[actions]